chore: add editorconfig

.editorconfig

@@ -0,0 +1,13 @@
+root = true
+
+[*]
+charset = utf-8
+end_of_line = crlf
+insert_final_newline = true
+trim_trailing_whitespace = true
+
+[*.md]
+trim_trailing_whitespace = false
+
+[*.{java,xml,yml,yaml,properties,json,sql,txt,ps1}]
+charset = utf-8

devflow/index.md

@@ -5,6 +5,7 @@
 | 日期 | slug | 领域 | 关键词 | 状态 |
 |---|---|---|---|---|
 | 2026-07-05 | diagnosis-playbook-skills | Agent Skill/Playbook | read_skill, diagnosis playbook, progressive disclosure, payment timeout, MySQL pool, Redis timeout | openspec/changes/diagnosis-playbook-skills | implemented |
+| 2026-07-07 | executor-evidence-output-contract | Chat质量门禁/证据归因 | Executor structured output, evidence bindings, Verifier structured claims, LOW_CONFID, hallucination | openspec/changes/archive/2026-07-07-executor-evidence-output-contract | archived |
 | 2026-07-06 | rag-eval-pipeline-closure | RAG/评测/回归闭环 | lookupResult fixture, LookupKnowledgeTool snapshot, evidenceBlocks, contextPack, retrievalTrace, rerankTrace, baseline diff, fallback case | devflow/projects/2026-07-06-rag-eval-pipeline-closure | archived |
 | 2026-07-06 | modular-rag-pipeline | RAG/Agent工具/证据链 | modular RAG, lookup_knowledge, evidenceBlocks, contextPack, rerank, retrievalTrace, L0 hint, unfiltered retry | openspec/changes/archive/2026-07-06-modular-rag-pipeline | archived |
 | 2026-07-05 | mvp-demo-interview-runbook | MVP Demo/Interview | Plan C, payment timeout, runbook, trace checklist, demo script | openspec/changes/archive/2026-07-05-mvp-demo-interview-runbook | archived |

devflow/projects/2026-07-07-executor-evidence-output-contract/acceptance.md

@@ -0,0 +1,16 @@
+# Acceptance: executor-evidence-output-contract
+
+## Draft Acceptance
+
+- [x] Issue exists: `mvp/issues/executor-evidence-attribution-hallucination.md`.
+- [x] OpenSpec change artifacts exist.
+- [x] devflow tracking files exist.
+- [x] OpenSpec validation passes.
+- [x] Implementation updates Chat Executor prompt.
+- [x] Implementation passes structured Executor output to Verifier.
+- [x] Verifier prefers structured claims and still falls back safely.
+- [x] Focused tests cover parsing, payload assembly, and unsupported confirmed claims.
+
+## Notes
+
+This project is currently in proposal/design stage. Runtime code is intentionally not changed yet.

devflow/projects/2026-07-07-executor-evidence-output-contract/brief.md

@@ -0,0 +1,23 @@
+# Brief: executor-evidence-output-contract
+
+## Summary
+
+Executor currently returns natural-language diagnosis answers that may mix confirmed evidence, runbook guidance, historical patterns, and unsupported inference. Verifier catches many unsupported facts, but only after extracting claims from prose.
+
+This project defines a structured Executor evidence-attribution contract and updates the Verifier input/verification path to consume it.
+
+## Goal
+
+Make Chat Executor output machine-checkable so confirmed claims are explicitly bound to current-session evidence, while hypotheses and evidence gaps remain visibly separate.
+
+## Scope
+
+- Chat Executor prompt contract.
+- Executor structured output parsing.
+- Verifier payload extension.
+- Chat Verifier prompt behavior.
+- Focused tests/eval fixtures.
+
+## Related OpenSpec
+
+`openspec/changes/executor-evidence-output-contract/`

devflow/projects/2026-07-07-executor-evidence-output-contract/decisions.md

@@ -0,0 +1,71 @@
+# Decisions: executor-evidence-output-contract
+
+## sm-flow Progress
+
+### Clarify
+
+Entry summary: recent Chat diagnosis sessions are `LOW_CONFID` because Executor presents unsupported or weakly supported details as confirmed facts after successful tool calls.
+
+Slug: `executor-evidence-output-contract`
+
+Scale: standard. This affects prompts, verifier input assembly, parsing behavior, and tests, but does not require a database schema change.
+
+### Context
+
+Relevant history:
+
+- `executor-action-memory-relevance`: Executor already has retrieval quality constraints and should avoid repeated `lookup_knowledge`.
+- `chat-verifier-agent`: Verifier should not see intermediate reasoning; it receives explicit `original_query`, `executor_final_answer`, `tool_trace_summary`, and `retry_context`.
+- `evidence-trace-hardening`: evidence-bearing tools persist stable traces and no-evidence semantics.
+- `modular-rag-pipeline`: `lookup_knowledge` exposes evidence blocks and context packs; L0 hints are not fact evidence.
+
+Current code shape:
+
+- `src/main/resources/prompts/chat-executor-prompt.md` is the Chat Executor prompt.
+- `src/main/resources/prompts/executor-prompt.md` is for the AiOps flow and is not the target of this Chat change.
+- `VerifierInputHook` currently builds a payload with `original_query`, `executor_final_answer`, `tool_trace_summary`, and `retry_context`.
+- Verifier prompt currently extracts facts from `executor_final_answer`.
+
+### Grill
+
+Question: Should Executor output only JSON or JSON plus readable answer?
+
+Decision: use one JSON object containing both machine fields and `user_facing_answer`. This avoids losing a readable Chinese answer while giving Verifier structured claims.
+
+Question: Should evidence binding use `chunk_id`?
+
+Decision: no. Use generic binding fields because `query_logs` and `query_metrics` do not naturally expose RAG chunks.
+
+Question: Should Verifier trust Executor-provided claims completely?
+
+Decision: no. Verifier should verify structured claims first, then scan `user_facing_answer` for extra confirmed-sounding facts omitted from `claims`.
+
+Question: What happens when Executor JSON is malformed?
+
+Decision: preserve raw final answer, mark parse failure, and fall back to existing natural-language verification.
+
+### Specify
+
+OpenSpec artifacts:
+
+- `proposal.md`: why and scope
+- `design.md`: contract, verifier behavior, risks
+- `specs/chat-verifier-agent/spec.md`: modified and added requirements
+- `tasks.md`: implementation checklist
+
+### Audit
+
+Cross-artifact alignment:
+
+- Issue describes evidence attribution hallucination.
+- Proposal scopes the fix to Executor output and Verifier consumption.
+- Design preserves existing verifier isolation.
+- Spec adds observable behavior without changing database schema.
+- Tasks remain implementation-oriented and unchecked.
+
+Interface impact:
+
+- Prompt/output contract: L2 internal Agent contract change.
+- Verifier payload: L2 internal structured input extension.
+- Database schema: no change.
+- External HTTP API: no intended change.

devflow/projects/2026-07-07-executor-evidence-output-contract/evidence.md

@@ -0,0 +1,17 @@
+# Evidence: executor-evidence-output-contract
+
+## Repository Evidence
+
+- `chat-executor-prompt.md` currently requires using real tool data but does not require a structured evidence-attribution output.
+- `chat-verifier-prompt.md` currently extracts facts from `executor_final_answer` prose and compares them with `tool_trace_summary`.
+- `VerifierInputHook` currently provides `original_query`, `executor_final_answer`, `tool_trace_summary`, and `retry_context`.
+- `openspec/specs/chat-verifier-agent/spec.md` already requires explicit verifier inputs, auditable evidence refs, fixed verdicts, and low-confidence handling.
+- `openspec/specs/evidence-trace-hardening/spec.md` already distinguishes failed, no-hit, deduped, and successful evidence-tool traces.
+
+## Runtime Evidence From Recent Sessions
+
+Recent MySQL inspection showed repeated `LOW_CONFID` verifier results with many `no_evidence` facts. Typical unsupported claims included OOM, Full GC frequency, specific slow SQL timings, lock waits, and service-specific timeout details that were not supported by current-session tool traces.
+
+## Design Evidence
+
+This change preserves the previous design that Verifier should not inspect intermediate reasoning. The new structured output is still final Executor output, not hidden chain-of-thought.

mvp/eval/schema.md

@@ -58,6 +58,7 @@ trace fixture 是一次 Agent 运行后的“留痕快照”。评测器不会
 | `session.answer` | Agent 最终给用户的回答 | 用来检查根因关键词和禁用词 |
 | `session.totalDurationMs` | 这次运行耗时 | 进入报告，帮助观察性能是否明显变差 |
 | `session.selfEvaluation.verifier_evaluation.verdict` | Verifier 对最终回答的判断 | 必须存在，并且要落在 case 的 `allowedVerdicts` 里 |
+| `session.selfEvaluation.verifier_evaluation.executor_structured_output.claims[*].evidence_bindings` | Executor 结构化输出里的确认事实证据绑定 | 如果 trace 中存在结构化 Executor 输出，每条 confirmed claim 必须有证据绑定 |
 | `session.selfEvaluation.verifier_evaluation.tool_trace_summary[*].tool_name` | Verifier 总结里看到的工具证据 | 用来补充判断证据工具是否出现 |
 | `toolInvocations[*].toolName` | Agent 实际调用过的工具名 | 用来检查 `requiredEvidenceTools` 是否满足 |
 | `toolInvocations[*].success` | 工具调用是否成功 | 当前主要保留在 trace 里，后续可以升级成更严格的成功率检查 |

openspec/changes/archive/2026-07-07-executor-evidence-output-contract/design.md

@@ -0,0 +1,161 @@
+## Context
+
+Current Chat flow explicitly runs `Planner -> Executor -> Verifier`. `VerifierInputHook` builds a structured verifier payload containing:
+
+- `original_query`
+- `executor_final_answer`
+- `tool_trace_summary`
+- `retry_context`
+
+This satisfies the earlier design goal that Verifier should not see Planner/Executor intermediate reasoning. However, Executor's answer is still natural language. Verifier must infer claims from prose, then compare those claims to tool traces. Recent low-confidence sessions show that Executor often converts weak or unrelated evidence into confirmed conclusions before Verifier sees it.
+
+The fix should move evidence attribution earlier: Executor must state which claims are confirmed, which evidence supports them, and which points remain hypotheses or gaps.
+
+## Goals / Non-Goals
+
+**Goals:**
+
+- Make Executor final output machine-checkable.
+- Separate confirmed claims, hypotheses, recommended actions, and missing information.
+- Make every confirmed claim bind to concrete tool evidence identifiers or excerpts.
+- Let Verifier consume structured claims directly instead of reconstructing all facts from prose.
+- Preserve Verifier input isolation from intermediate reasoning.
+- Preserve existing behavior when Executor output is malformed by falling back to natural-language verification and low-confidence handling.
+
+**Non-Goals:**
+
+- Do not add database tables or columns.
+- Do not change evidence tool method signatures.
+- Do not let Verifier call tools.
+- Do not require full raw tool outputs in verifier input.
+- Do not make user-facing answers become raw JSON-only unless the product layer explicitly chooses that rendering.
+
+## Decisions
+
+### Decision: Executor emits an evidence-attribution contract
+
+Executor final output SHALL be a single JSON object. The recommended contract is:
+
+```json
+{
+  "answer_version": "executor_evidence_v1",
+  "diagnosis_summary": "1-2 sentence summary using only supported facts",
+  "claims": [
+    {
+      "claim_id": "claim-1",
+      "claim_type": "root_cause",
+      "claim_text": "The payment-service connection pool is saturated",
+      "support_level": "direct",
+      "evidence_bindings": [
+        {
+          "source_type": "tool_trace",
+          "source_id": "trace-1",
+          "tool_name": "query_metrics",
+          "source_invocation_ids": [101],
+          "evidence_excerpt": "active connections reached max pool size"
+        }
+      ]
+    }
+  ],
+  "hypotheses": [
+    {
+      "hypothesis_text": "A connection leak may be contributing",
+      "basis": "metrics show saturation, but no leak evidence was returned",
+      "needed_evidence": ["connection lifetime metrics", "leak detection logs"]
+    }
+  ],
+  "recommended_actions": [
+    {
+      "action_text": "Check HikariCP active/idle/pending connection metrics",
+      "reason": "Needed to confirm pool saturation scope",
+      "evidence_bindings": []
+    }
+  ],
+  "missing_info": [
+    "No log evidence confirming a connection leak was returned"
+  ],
+  "user_facing_answer": "Chinese answer rendered from the same confirmed claims, hypotheses, actions, and gaps"
+}
+```
+
+Rationale: this keeps the machine contract explicit while still allowing the product to return a readable Chinese answer.
+
+### Decision: Evidence binding uses generic source fields
+
+`chunk_id` alone is too specific to `lookup_knowledge`. The binding shape SHALL support all evidence-bearing tools:
+
+- `source_type`: `tool_trace`, `lookup_evidence_block`, or another stable source family
+- `source_id`: `tool_trace_summary.trace_ref`, evidence block id, or equivalent stable id
+- `tool_name`: evidence tool name when available
+- `source_invocation_ids`: persisted `tool_invocation` ids when available
+- `evidence_excerpt`: bounded excerpt copied or summarized from tool evidence
+
+Rationale: `query_logs`, `query_metrics`, and `lookup_knowledge` expose evidence differently. A generic binding prevents the prompt contract from overfitting to RAG chunks.
+
+### Decision: Confirmed claims are stricter than hypotheses and actions
+
+Confirmed `claims` SHALL contain only facts supported by current-session evidence. Runbook instructions, skill guidance, and historical case patterns SHALL NOT appear as confirmed claims unless the current tool trace supports that exact fact.
+
+Unsupported but useful diagnostic ideas SHALL be placed under `hypotheses` or `recommended_actions`.
+
+Rationale: this directly addresses the observed hallucination: turning plausible patterns into current incident facts.
+
+### Decision: Verifier prefers structured claims but keeps fallback
+
+Verifier prompt SHALL use this order:
+
+1. If `executor_structured_output.claims` is valid, verify each claim directly.
+2. Also scan `user_facing_answer` for extra confirmed-sounding facts not present in `claims`; mark them as facts to check.
+3. If the structured output is missing or invalid, fall back to the existing natural-language extraction from `executor_final_answer`.
+
+Invalid structured output SHALL NOT crash the flow. It SHOULD produce a fallback verifier decision and make malformed structure visible in `verifier_evaluation`.
+
+Rationale: the new contract should improve precision without making runtime brittle.
+
+### Decision: Verifier still does not see intermediate reasoning
+
+The verifier payload MAY add:
+
+- `executor_structured_output`
+- `executor_output_parse_status`
+
+It SHALL continue to exclude Planner reasoning, Executor intermediate reasoning, and raw conversation noise.
+
+Rationale: this preserves the original verifier design: verify the final answer and tool traces, not hidden reasoning.
+
+### Decision: User-facing output remains Chinese
+
+For Chat user-facing pages and answers, the rendered answer SHALL be Chinese. If the Executor emits JSON, either:
+
+- `user_facing_answer` is returned to the user after verifier routing, or
+- ChatService renders a Chinese answer from the structured contract.
+
+The raw machine contract may remain visible in trace/debug views, but the normal user answer should not become an English/JSON-only artifact.
+
+Rationale: this aligns with current UI/product language requirements while keeping machine-checkable evidence.
+
+## Risks / Trade-offs
+
+- [Risk] LLM may emit malformed JSON. Mitigation: parse-status fallback and verifier natural-language fallback.
+- [Risk] Executor may put unsupported facts in `user_facing_answer` but omit them from `claims`. Mitigation: Verifier scans `user_facing_answer` for extra confirmed-sounding facts.
+- [Risk] Evidence excerpts may be fabricated. Mitigation: Verifier checks binding ids against `tool_trace_summary` and treats missing refs as no evidence.
+- [Risk] More verbose output increases tokens. Mitigation: keep excerpts bounded and put detailed raw evidence only in tool traces.
+- [Risk] Prompt-only enforcement is soft. Mitigation: add focused tests/eval fixtures and later consider code-level schema validation.
+
+## Migration Plan
+
+1. Update `chat-executor-prompt.md` with the evidence-attribution JSON contract.
+2. Add parsing in Chat runtime for Executor output:
+   - valid JSON -> preserve parsed `executor_structured_output`
+   - invalid JSON -> mark parse status and keep raw `executor_final_answer`
+3. Update verifier payload assembly to include `executor_structured_output` and parse status.
+4. Update `chat-verifier-prompt.md` to prefer structured claims and verify extra facts in `user_facing_answer`.
+5. Persist parsed or raw structured output in existing trace/evaluation snapshots without schema changes.
+6. Add focused tests and regression fixtures for unsupported confirmed claims.
+7. Validate OpenSpec and run targeted tests.
+
+## Open Questions
+
+- Should `user_facing_answer` be mandatory in Executor JSON, or should ChatService render it from structured fields?
+- Should malformed Executor JSON force `LOW_CONFID`, or should Verifier decide based on natural-language fallback alone?
+- Should `support_level` allow only `direct`, `indirect`, and `none`, or should it also include `contradicted` for Executor self-reporting?

openspec/changes/archive/2026-07-07-executor-evidence-output-contract/proposal.md

@@ -0,0 +1,42 @@
+# Proposal: executor-evidence-output-contract
+
+## Why
+
+Recent Chat diagnosis sessions frequently end as `LOW_CONFID` even though evidence tools were called successfully. The main failure mode is not missing tool execution; it is that Executor blends tool evidence, runbook/reference patterns, and model inference into one natural-language answer, then presents unsupported details as confirmed incident facts.
+
+Verifier currently receives `original_query`, `executor_final_answer`, `tool_trace_summary`, and optional `retry_context`. It can catch unsupported claims, but it must first infer facts from unstructured prose. That makes the quality gate reactive and noisy: unsupported claims are detected after the answer has already been shaped as a confident story.
+
+## What Changes
+
+- Update the Chat Executor prompt so its final output follows a strict evidence-attribution JSON contract.
+- Require confirmed claims to carry explicit evidence bindings and require unsupported items to be placed in `hypotheses`, `missing_info`, or `recommended_actions` instead of confirmed conclusions.
+- Extend the verifier input contract with a parsed or raw `executor_structured_output` field while preserving the existing `executor_final_answer` field for fallback compatibility.
+- Update the Chat Verifier prompt so it prefers Executor-provided structured claims over natural-language fact extraction.
+- Keep Verifier isolated from Planner/Executor intermediate reasoning; it still receives only the original query, Executor final output, structured claim contract, tool trace summary, and retry context.
+- Do not change evidence tool signatures, database schema, or the `Planner -> Executor -> Verifier` topology.
+
+## Capabilities
+
+### New Capabilities
+
+None.
+
+### Modified Capabilities
+
+- `chat-verifier-agent`: Chat verification now consumes an Executor evidence-attribution contract when available.
+
+## Impact
+
+- Affected prompts: `chat-executor-prompt.md`, `chat-verifier-prompt.md`.
+- Affected integration: `VerifierInputHook` or equivalent verifier payload assembly must include `executor_structured_output` when Executor returns valid structured JSON.
+- Affected parsing: `ChatService` may parse Executor output to separate machine contract from user-facing text, with fallback when parsing fails.
+- Affected tests/eval: prompt contract tests, verifier input assembly tests, and unsupported-claim regression fixtures.
+- No database schema change is required. Structured Executor output can be persisted in existing step output fields and verifier snapshots.
+
+## Out Of Scope
+
+- Loosening Verifier scoring or `PASS` criteria.
+- Raising `verifier.low-confidence-threshold` to hide unsupported claims.
+- Treating runbook, skill, or historical case text as current incident evidence unless it was returned as evidence for the current query and bound explicitly.
+- Adding a new verifier tool or allowing Verifier to perform retrieval.
+- Replacing the existing tool trace summary contract.

openspec/changes/archive/2026-07-07-executor-evidence-output-contract/specs/chat-verifier-agent/spec.md

@@ -0,0 +1,95 @@
+## MODIFIED Requirements
+
+### Requirement: Verifier SHALL consume explicit verification inputs
+The Verifier SHALL receive explicit verification inputs rather than inferring them only from raw conversation history.
+
+#### Scenario: explicit input blocks available to Verifier
+- **WHEN** the Verifier starts
+- **THEN** the system SHALL provide `original_query`, `executor_final_answer`, and `tool_trace_summary` as explicit inputs
+- **AND** `retry_context` SHALL be provided on the second round only
+- **AND** when Executor returns a valid evidence-attribution contract, the system SHALL provide `executor_structured_output`
+- **AND** when Executor output parsing fails, the system SHALL provide an `executor_output_parse_status` that indicates the failure
+- **AND** message filtering MAY be used only to remove intermediate reasoning or unrelated noise
+
+#### Scenario: Verifier remains isolated from intermediate reasoning
+- **WHEN** `executor_structured_output` is added to the verifier input
+- **THEN** the input SHALL still exclude Planner reasoning and Executor intermediate reasoning
+- **AND** the input SHALL be limited to the original query, final Executor output, parsed Executor evidence contract, tool trace summary, and retry context
+
+### Requirement: Verifier SHALL fact-check Executor answers
+The system SHALL have a Verifier Agent that reads the Executor's answer and the tool call history, then produces a structured verdict.
+
+#### Scenario: Structured Executor claims are verified first
+- **WHEN** `executor_structured_output.claims` is present and valid
+- **THEN** Verifier SHALL verify each structured claim against `tool_trace_summary`
+- **AND** each claim's evidence bindings SHALL reference existing trace or invocation identifiers when those identifiers are available
+- **AND** a claim with fabricated or missing evidence references SHALL NOT be classified as `direct_evidence`
+
+#### Scenario: Extra confirmed-sounding answer facts are still checked
+- **WHEN** `executor_structured_output.user_facing_answer` contains confirmed-sounding facts that are absent from `executor_structured_output.claims`
+- **THEN** Verifier SHALL add those facts to `facts_checked`
+- **AND** unsupported extra facts SHALL lower the verdict according to the existing verdict matrix
+
+#### Scenario: Natural-language fallback remains available
+- **WHEN** Executor does not return parseable structured output
+- **THEN** Verifier SHALL fall back to extracting facts from `executor_final_answer`
+- **AND** the final verdict SHALL still follow the existing groundedness and evidence classification rules
+
+## ADDED Requirements
+
+### Requirement: Executor SHALL output an evidence-attribution contract
+The Chat Executor SHALL produce a machine-checkable final output that separates confirmed claims from hypotheses, recommendations, and missing information.
+
+#### Scenario: Executor final output contains required top-level fields
+- **WHEN** Executor completes a Chat diagnosis step
+- **THEN** its final output SHALL contain `answer_version`, `diagnosis_summary`, `claims`, `hypotheses`, `recommended_actions`, `missing_info`, and `user_facing_answer`
+- **AND** the output SHOULD be parseable as one JSON object without Markdown fences
+
+#### Scenario: Confirmed claims carry evidence bindings
+- **WHEN** Executor emits an item under `claims`
+- **THEN** the item SHALL include `claim_id`, `claim_type`, `claim_text`, `support_level`, and `evidence_bindings`
+- **AND** `support_level` SHALL be one of `direct`, `indirect`, or `none`
+- **AND** claims with `support_level=direct` or `support_level=indirect` SHALL include at least one evidence binding
+
+#### Scenario: Evidence bindings support multiple tool types
+- **WHEN** Executor binds evidence to a claim
+- **THEN** each binding SHALL include `source_type`, `source_id`, `tool_name`, `source_invocation_ids`, and `evidence_excerpt`
+- **AND** the binding SHALL be able to reference `lookup_knowledge`, `query_logs`, `query_metrics`, or other evidence-bearing tool traces
+- **AND** the binding SHALL NOT rely only on a RAG-specific `chunk_id`
+
+#### Scenario: Unsupported conclusions are not confirmed claims
+- **WHEN** a possible root cause, detail, or remediation lacks current-session tool evidence
+- **THEN** Executor SHALL place it under `hypotheses`, `recommended_actions`, or `missing_info`
+- **AND** Executor SHALL NOT present it as a confirmed claim
+
+#### Scenario: Runbook and skill guidance do not become incident facts
+- **WHEN** Executor uses runbook, skill, or historical-case guidance
+- **THEN** the guidance MAY influence `recommended_actions`
+- **AND** the guidance SHALL NOT be emitted as a current incident fact unless current-session tool evidence supports it
+
+### Requirement: User-facing Chat answers SHALL remain readable Chinese
+The system SHALL preserve a readable Chinese answer for normal Chat users even when Executor emits a machine-checkable contract.
+
+#### Scenario: User-facing answer is available
+- **WHEN** Executor emits structured output
+- **THEN** `user_facing_answer` SHALL be written in Chinese
+- **AND** it SHALL be consistent with the confirmed claims, hypotheses, recommended actions, and missing information in the same JSON object
+
+#### Scenario: Machine contract remains available for trace inspection
+- **WHEN** the Chat trace or verifier evaluation is inspected
+- **THEN** the structured Executor contract MAY be shown for debugging or audit
+- **AND** normal user output SHALL use the existing verifier-routed display path rather than exposing raw JSON by default
+
+### Requirement: Structured Executor output SHALL degrade safely
+The system SHALL tolerate malformed or absent structured Executor output without crashing the Chat flow.
+
+#### Scenario: Malformed Executor JSON is captured
+- **WHEN** Executor returns malformed JSON or text outside the expected contract
+- **THEN** Chat runtime SHALL preserve the raw `executor_final_answer`
+- **AND** it SHALL mark `executor_output_parse_status` as failed
+- **AND** Verifier SHALL use natural-language fallback behavior
+
+#### Scenario: Structured parse failure remains observable
+- **WHEN** Executor output parsing fails
+- **THEN** the verifier evaluation or trace snapshot SHALL make the parse failure visible
+- **AND** the failure SHALL NOT be silently treated as a successful evidence-attribution contract

openspec/changes/archive/2026-07-07-executor-evidence-output-contract/tasks.md

@@ -0,0 +1,31 @@
+## 1. Executor Contract
+
+- [x] 1.1 Update `chat-executor-prompt.md` with the evidence-attribution JSON contract.
+- [x] 1.2 Require confirmed claims, hypotheses, recommended actions, missing information, and Chinese `user_facing_answer`.
+- [x] 1.3 Add prompt-level constraints that runbook/skill/history guidance cannot become current incident facts without current-session evidence.
+
+## 2. Runtime Parsing And Verifier Input
+
+- [x] 2.1 Parse Executor final output into `executor_structured_output` when it is valid JSON.
+- [x] 2.2 Add `executor_output_parse_status` for valid, missing, and malformed outputs.
+- [x] 2.3 Pass `executor_structured_output` and parse status into the verifier payload while preserving existing `executor_final_answer`.
+- [x] 2.4 Persist or expose the structured output through existing trace/evaluation snapshots without a schema change.
+
+## 3. Verifier Prompt
+
+- [x] 3.1 Update `chat-verifier-prompt.md` to verify structured claims before natural-language extraction.
+- [x] 3.2 Add a verifier rule to scan `user_facing_answer` for extra confirmed-sounding facts not present in structured `claims`.
+- [x] 3.3 Keep the existing natural-language fallback for malformed or absent structured output.
+
+## 4. Tests And Evaluation
+
+- [x] 4.1 Add focused tests for Executor output parsing and parse-status fallback.
+- [x] 4.2 Add focused tests for verifier payload assembly with structured Executor output.
+- [x] 4.3 Add prompt/fixture regression coverage for unsupported claims being placed outside confirmed `claims`.
+- [x] 4.4 Add or update eval fixture checks that confirmed claims must have evidence bindings.
+
+## 5. Verification
+
+- [x] 5.1 Run targeted unit tests for the parsing and verifier-input changes.
+- [x] 5.2 Run relevant Chat verifier/eval regression tests.
+- [x] 5.3 Run OpenSpec validation for `executor-evidence-output-contract`.

openspec/specs/chat-verifier-agent/spec.md

@@ -27,6 +27,22 @@ The system SHALL have a Verifier Agent that reads the Executor's answer and the
 - **OR** the answer fabricates a key entity, error code, or conclusion that does not exist in the tool evidence
 - **THEN** the Verifier SHALL output verdict="REJECT"
 
+#### Scenario: Structured Executor claims are verified first
+- **WHEN** `executor_structured_output.claims` is present and valid
+- **THEN** Verifier SHALL verify each structured claim against `tool_trace_summary`
+- **AND** each claim's evidence bindings SHALL reference existing trace or invocation identifiers when those identifiers are available
+- **AND** a claim with fabricated or missing evidence references SHALL NOT be classified as `direct_evidence`
+
+#### Scenario: Extra confirmed-sounding answer facts are still checked
+- **WHEN** `executor_structured_output.user_facing_answer` contains confirmed-sounding facts that are absent from `executor_structured_output.claims`
+- **THEN** Verifier SHALL add those facts to `facts_checked`
+- **AND** unsupported extra facts SHALL lower the verdict according to the existing verdict matrix
+
+#### Scenario: Natural-language fallback remains available
+- **WHEN** Executor does not return parseable structured output
+- **THEN** Verifier SHALL fall back to extracting facts from `executor_final_answer`
+- **AND** the final verdict SHALL still follow the existing groundedness and evidence classification rules
+
 ### Requirement: Verifier SHALL output structured JSON
 The Verifier SHALL output a JSON object with verdict, groundedness_score, facts_checked array, and rationale.
 
@@ -138,8 +154,15 @@ The Verifier SHALL receive explicit verification inputs rather than inferring th
 - **WHEN** the Verifier starts
 - **THEN** the system SHALL provide `original_query`, `executor_final_answer`, and `tool_trace_summary` as explicit inputs
 - **AND** `retry_context` SHALL be provided on the second round only
+- **AND** when Executor returns a valid evidence-attribution contract, the system SHALL provide `executor_structured_output`
+- **AND** when Executor output parsing fails, the system SHALL provide an `executor_output_parse_status` that indicates the failure
 - **AND** message filtering MAY be used only to remove intermediate reasoning or unrelated noise
 
+#### Scenario: Verifier remains isolated from intermediate reasoning
+- **WHEN** `executor_structured_output` is added to the verifier input
+- **THEN** the input SHALL still exclude Planner reasoning and Executor intermediate reasoning
+- **AND** the input SHALL be limited to the original query, final Executor output, parsed Executor evidence contract, tool trace summary, and retry context
+
 #### Scenario: tool trace summary derived from tool facts
 - **WHEN** the system prepares verifier inputs
 - **THEN** `tool_trace_summary` SHALL be generated from tool invocation facts
@@ -230,3 +253,59 @@ When Verifier returns `LOW_CONFID`, user-facing output SHALL clearly separate co
 - **THEN** the final user-facing response SHALL identify those gaps from verifier output
 - **AND** unsupported raw Executor claims SHALL NOT be presented as confirmed conclusions
 
+### Requirement: Executor SHALL output an evidence-attribution contract
+The Chat Executor SHALL produce a machine-checkable final output that separates confirmed claims from hypotheses, recommendations, and missing information.
+
+#### Scenario: Executor final output contains required top-level fields
+- **WHEN** Executor completes a Chat diagnosis step
+- **THEN** its final output SHALL contain `answer_version`, `diagnosis_summary`, `claims`, `hypotheses`, `recommended_actions`, `missing_info`, and `user_facing_answer`
+- **AND** the output SHOULD be parseable as one JSON object without Markdown fences
+
+#### Scenario: Confirmed claims carry evidence bindings
+- **WHEN** Executor emits an item under `claims`
+- **THEN** the item SHALL include `claim_id`, `claim_type`, `claim_text`, `support_level`, and `evidence_bindings`
+- **AND** `support_level` SHALL be one of `direct`, `indirect`, or `none`
+- **AND** claims with `support_level=direct` or `support_level=indirect` SHALL include at least one evidence binding
+
+#### Scenario: Evidence bindings support multiple tool types
+- **WHEN** Executor binds evidence to a claim
+- **THEN** each binding SHALL include `source_type`, `source_id`, `tool_name`, `source_invocation_ids`, and `evidence_excerpt`
+- **AND** the binding SHALL be able to reference `lookup_knowledge`, `query_logs`, `query_metrics`, or other evidence-bearing tool traces
+- **AND** the binding SHALL NOT rely only on a RAG-specific `chunk_id`
+
+#### Scenario: Unsupported conclusions are not confirmed claims
+- **WHEN** a possible root cause, detail, or remediation lacks current-session tool evidence
+- **THEN** Executor SHALL place it under `hypotheses`, `recommended_actions`, or `missing_info`
+- **AND** Executor SHALL NOT present it as a confirmed claim
+
+#### Scenario: Runbook and skill guidance do not become incident facts
+- **WHEN** Executor uses runbook, skill, or historical-case guidance
+- **THEN** the guidance MAY influence `recommended_actions`
+- **AND** the guidance SHALL NOT be emitted as a current incident fact unless current-session tool evidence supports it
+
+### Requirement: User-facing Chat answers SHALL remain readable Chinese
+The system SHALL preserve a readable Chinese answer for normal Chat users even when Executor emits a machine-checkable contract.
+
+#### Scenario: User-facing answer is available
+- **WHEN** Executor emits structured output
+- **THEN** `user_facing_answer` SHALL be written in Chinese
+- **AND** it SHALL be consistent with the confirmed claims, hypotheses, recommended actions, and missing information in the same JSON object
+
+#### Scenario: Machine contract remains available for trace inspection
+- **WHEN** the Chat trace or verifier evaluation is inspected
+- **THEN** the structured Executor contract MAY be shown for debugging or audit
+- **AND** normal user output SHALL use the existing verifier-routed display path rather than exposing raw JSON by default
+
+### Requirement: Structured Executor output SHALL degrade safely
+The system SHALL tolerate malformed or absent structured Executor output without crashing the Chat flow.
+
+#### Scenario: Malformed Executor JSON is captured
+- **WHEN** Executor returns malformed JSON or text outside the expected contract
+- **THEN** Chat runtime SHALL preserve the raw `executor_final_answer`
+- **AND** it SHALL mark `executor_output_parse_status` as failed
+- **AND** Verifier SHALL use natural-language fallback behavior
+
+#### Scenario: Structured parse failure remains observable
+- **WHEN** Executor output parsing fails
+- **THEN** the verifier evaluation or trace snapshot SHALL make the parse failure visible
+- **AND** the failure SHALL NOT be silently treated as a successful evidence-attribution contract

src/main/java/com/superbiz/agent/controller/FaviconController.java

@@ -0,0 +1,19 @@
+package com.superbiz.agent.controller;
+
+import org.springframework.http.CacheControl;
+import org.springframework.http.ResponseEntity;
+import org.springframework.stereotype.Controller;
+import org.springframework.web.bind.annotation.GetMapping;
+
+import java.time.Duration;
+
+@Controller
+public class FaviconController {
+
+    @GetMapping("/favicon.ico")
+    public ResponseEntity<Void> favicon() {
+        return ResponseEntity.noContent()
+                .cacheControl(CacheControl.maxAge(Duration.ofDays(1)))
+                .build();
+    }
+}

src/main/java/com/superbiz/agent/eval/DiagnosisTraceEvaluator.java

@@ -101,6 +101,8 @@ public class DiagnosisTraceEvaluator {
             failedChecks.add("reject output does not use degraded template");
         }
 
+        failedChecks.addAll(validateExecutorStructuredOutput(trace));
+
         Integer toolCallCount = trace.getToolInvocations() == null ? 0 : trace.getToolInvocations().size();
         Integer durationMs = trace.getSession() == null ? null : trace.getSession().getTotalDurationMs();
 
@@ -174,6 +176,32 @@ public class DiagnosisTraceEvaluator {
         return value == null ? null : String.valueOf(value);
     }
 
+    private List<String> validateExecutorStructuredOutput(DiagnosisTraceResponse trace) {
+        Object structuredOutput = nestedValue(trace, "verifier_evaluation", "executor_structured_output");
+        if (!(structuredOutput instanceof Map<?, ?> output)) {
+            return List.of();
+        }
+        Object claims = output.get("claims");
+        if (!(claims instanceof List<?> claimList)) {
+            return List.of("executor structured output missing claims array");
+        }
+
+        List<String> failedChecks = new ArrayList<>();
+        for (Object item : claimList) {
+            if (!(item instanceof Map<?, ?> claim)) {
+                failedChecks.add("executor structured claim is not an object");
+                continue;
+            }
+            Object claimIdValue = claim.get("claim_id");
+            String claimId = claimIdValue == null ? "unknown" : String.valueOf(claimIdValue);
+            Object bindings = claim.get("evidence_bindings");
+            if (!(bindings instanceof List<?> bindingList) || bindingList.isEmpty()) {
+                failedChecks.add("executor confirmed claim missing evidence bindings: " + claimId);
+            }
+        }
+        return failedChecks;
+    }
+
     private Object nestedValue(DiagnosisTraceResponse trace, String firstKey, String secondKey) {
         if (trace.getSession() == null || trace.getSession().getSelfEvaluation() == null) {
             return null;

src/main/java/com/superbiz/agent/hook/VerifierInputHook.java

@@ -5,6 +5,8 @@ import com.alibaba.cloud.ai.graph.agent.hook.HookPosition;
 import com.alibaba.cloud.ai.graph.agent.hook.HookPositions;
 import com.alibaba.cloud.ai.graph.agent.hook.messages.AgentCommand;
 import com.alibaba.cloud.ai.graph.agent.hook.messages.MessagesModelHook;
+import com.fasterxml.jackson.core.type.TypeReference;
+import com.fasterxml.jackson.databind.JsonNode;
 import com.fasterxml.jackson.databind.ObjectMapper;
 import com.superbiz.agent.service.ToolTraceSummaryService;
 import com.superbiz.agent.util.SessionContextHolder;
@@ -27,6 +29,8 @@ public class VerifierInputHook extends MessagesModelHook {
 
     private final ToolTraceSummaryService toolTraceSummaryService;
     private final ObjectMapper objectMapper = new ObjectMapper();
+    private static final TypeReference<Map<String, Object>> MAP_TYPE = new TypeReference<>() {
+    };
 
     public VerifierInputHook(ToolTraceSummaryService toolTraceSummaryService) {
         this.toolTraceSummaryService = toolTraceSummaryService;
@@ -48,6 +52,10 @@ public class VerifierInputHook extends MessagesModelHook {
                 executorFinalAnswer = extractLastAssistantText(previousMessages);
             }
 
+            ExecutorOutputParseResult parseResult = parseExecutorOutput(executorFinalAnswer);
+            VerifierContextHolder.setExecutorStructuredOutput(parseResult.structuredOutput());
+            VerifierContextHolder.setExecutorOutputParseStatus(parseResult.status());
+
             List<Map<String, Object>> toolTraceSummary =
                     toolTraceSummaryService.buildVerifierTraceSummary(sessionId, executorFinalAnswer);
             VerifierContextHolder.setToolTraceSummary(toolTraceSummary);
@@ -55,6 +63,8 @@ public class VerifierInputHook extends MessagesModelHook {
             Map<String, Object> verifierInput = new LinkedHashMap<>();
             verifierInput.put("original_query", VerifierContextHolder.getOriginalQuery());
             verifierInput.put("executor_final_answer", executorFinalAnswer);
+            verifierInput.put("executor_structured_output", parseResult.structuredOutput());
+            verifierInput.put("executor_output_parse_status", parseResult.status());
             verifierInput.put("tool_trace_summary", toolTraceSummary);
             verifierInput.put("retry_context", VerifierContextHolder.getRetryContext());
 
@@ -66,6 +76,59 @@ public class VerifierInputHook extends MessagesModelHook {
         }
     }
 
+    private ExecutorOutputParseResult parseExecutorOutput(String executorFinalAnswer) {
+        if (executorFinalAnswer == null || executorFinalAnswer.isBlank()) {
+            return new ExecutorOutputParseResult(null, status("missing", "executor_final_answer is blank"));
+        }
+
+        String sanitized = sanitizeJsonPayload(executorFinalAnswer);
+        if (!looksJsonLike(sanitized)) {
+            return new ExecutorOutputParseResult(null, status("missing", "executor output is not JSON"));
+        }
+
+        try {
+            JsonNode root = objectMapper.readTree(sanitized);
+            if (!root.isObject() || !root.path("claims").isArray()) {
+                return new ExecutorOutputParseResult(null, status("malformed",
+                        "executor output JSON does not match evidence-attribution contract"));
+            }
+            Map<String, Object> structuredOutput = objectMapper.convertValue(root, MAP_TYPE);
+            return new ExecutorOutputParseResult(structuredOutput, status("valid", "parsed executor evidence contract"));
+        } catch (Exception e) {
+            log.debug("Failed to parse executor structured output", e);
+            return new ExecutorOutputParseResult(null, status("malformed", e.getMessage()));
+        }
+    }
+
+    private String sanitizeJsonPayload(String raw) {
+        String trimmed = raw.trim();
+        int fenceStart = trimmed.indexOf("```");
+        if (fenceStart >= 0) {
+            int firstNewline = trimmed.indexOf('\n', fenceStart);
+            int lastFence = trimmed.indexOf("```", firstNewline + 1);
+            if (firstNewline >= 0 && lastFence > firstNewline) {
+                return trimmed.substring(firstNewline + 1, lastFence).trim();
+            }
+        }
+        int objectStart = trimmed.indexOf('{');
+        int objectEnd = trimmed.lastIndexOf('}');
+        if (objectStart >= 0 && objectEnd > objectStart) {
+            return trimmed.substring(objectStart, objectEnd + 1).trim();
+        }
+        return trimmed;
+    }
+
+    private boolean looksJsonLike(String text) {
+        return text.startsWith("{") && text.endsWith("}");
+    }
+
+    private Map<String, Object> status(String status, String detail) {
+        Map<String, Object> result = new LinkedHashMap<>();
+        result.put("status", status);
+        result.put("detail", detail == null ? "" : detail);
+        return result;
+    }
+
     private String extractLastAssistantText(List<Message> previousMessages) {
         for (int i = previousMessages.size() - 1; i >= 0; i--) {
             if (previousMessages.get(i) instanceof AssistantMessage assistantMessage) {
@@ -102,4 +165,10 @@ public class VerifierInputHook extends MessagesModelHook {
         }
         return message.toString();
     }
+
+    private record ExecutorOutputParseResult(
+            Map<String, Object> structuredOutput,
+            Map<String, Object> status
+    ) {
+    }
 }

src/main/java/com/superbiz/agent/repository/AgentStepRepository.java

@@ -17,6 +17,11 @@ public interface AgentStepRepository extends JpaRepository<AgentStep, Long> {
      */
     List<AgentStep> findBySessionIdOrderByStepIndex(String sessionId);
 
+    /**
+     * 根据会话ID查询所有步骤，调用方负责按具体视图语义排序。
+     */
+    List<AgentStep> findBySessionId(String sessionId);
+
     /**
      * 统计某个会话的步骤数
      */

src/main/java/com/superbiz/agent/service/ChatService.java

@@ -444,7 +444,8 @@ public class ChatService {
                 }
 
                 if ("PASS".equals(finalDecision.verdict())) {
-                    answer = answer == null || answer.isBlank() ? "抱歉，多 Agent 分析未能生成有效结论。" : answer;
+                    answer = extractUserFacingAnswer(answer)
+                            .orElse(answer == null || answer.isBlank() ? "抱歉，多 Agent 分析未能生成有效结论。" : answer);
                     persistVerifierEvaluation(session, finalDecision, round);
                     break;
                 }
@@ -744,6 +745,10 @@ public class ChatService {
         verifierEvaluation.put("rationale", decision.rationale());
         verifierEvaluation.put("round", round);
         verifierEvaluation.put("traceability_version", "v1");
+        verifierEvaluation.put("executor_output_parse_status",
+                Optional.ofNullable(VerifierContextHolder.getExecutorOutputParseStatus())
+                        .orElse(Map.of("status", "missing", "detail", "executor parse status unavailable")));
+        verifierEvaluation.put("executor_structured_output", VerifierContextHolder.getExecutorStructuredOutput());
         verifierEvaluation.put("tool_trace_summary",
                 Optional.ofNullable(VerifierContextHolder.getToolTraceSummary()).orElse(List.of()));
 
@@ -795,6 +800,22 @@ public class ChatService {
         return output.toString();
     }
 
+    private Optional<String> extractUserFacingAnswer(String executorAnswer) {
+        if (executorAnswer == null || executorAnswer.isBlank()) {
+            return Optional.empty();
+        }
+        try {
+            JsonNode root = objectMapper.readTree(sanitizeJsonPayload(executorAnswer));
+            String userFacingAnswer = root.path("user_facing_answer").asText("");
+            if (!userFacingAnswer.isBlank()) {
+                return Optional.of(userFacingAnswer);
+            }
+        } catch (Exception e) {
+            logger.debug("Executor answer is not structured JSON, keep raw answer");
+        }
+        return Optional.empty();
+    }
+
     private String buildDegradedOutput(VerifierDecision decision) {
         StringBuilder output = new StringBuilder(DEGRADED_PREFIX);
 

src/main/java/com/superbiz/agent/service/DiagnosisTraceService.java

@@ -13,6 +13,7 @@ import com.superbiz.agent.repository.ToolInvocationRepository;
 import lombok.RequiredArgsConstructor;
 import org.springframework.stereotype.Service;
 
+import java.util.Comparator;
 import java.util.List;
 import java.util.Map;
 
@@ -31,7 +32,7 @@ public class DiagnosisTraceService {
     public DiagnosisTraceResponse getTrace(String sessionId) {
         DiagnosisSession session = diagnosisSessionRepository.findBySessionId(sessionId)
                 .orElseThrow(() -> new SessionNotFoundException(sessionId));
-        List<AgentStep> steps = agentStepRepository.findBySessionIdOrderByStepIndex(sessionId);
+        List<AgentStep> steps = orderStepsForTrace(agentStepRepository.findBySessionId(sessionId));
         List<ToolInvocation> toolInvocations = toolInvocationRepository.findBySessionIdOrderByIdAsc(sessionId);
 
         return DiagnosisTraceResponse.builder()
@@ -42,6 +43,15 @@ public class DiagnosisTraceService {
                 .build();
     }
 
+    private List<AgentStep> orderStepsForTrace(List<AgentStep> steps) {
+        return steps.stream()
+                .sorted(Comparator
+                        .comparing(AgentStep::getCreatedAt, Comparator.nullsLast(Comparator.naturalOrder()))
+                        .thenComparing(AgentStep::getId, Comparator.nullsLast(Comparator.naturalOrder()))
+                        .thenComparing(AgentStep::getStepIndex, Comparator.nullsLast(Comparator.naturalOrder())))
+                .toList();
+    }
+
     private DiagnosisTraceResponse.SessionTrace toSessionTrace(DiagnosisSession session) {
         return DiagnosisTraceResponse.SessionTrace.builder()
                 .id(session.getId())

src/main/java/com/superbiz/agent/util/VerifierContextHolder.java

@@ -11,6 +11,8 @@ public final class VerifierContextHolder {
     private static final ThreadLocal<String> ORIGINAL_QUERY = new ThreadLocal<>();
     private static final ThreadLocal<String> RETRY_CONTEXT = new ThreadLocal<>();
     private static final ThreadLocal<String> EXECUTOR_FINAL_ANSWER = new ThreadLocal<>();
+    private static final ThreadLocal<Map<String, Object>> EXECUTOR_STRUCTURED_OUTPUT = new ThreadLocal<>();
+    private static final ThreadLocal<Map<String, Object>> EXECUTOR_OUTPUT_PARSE_STATUS = new ThreadLocal<>();
     private static final ThreadLocal<List<Map<String, Object>>> TOOL_TRACE_SUMMARY = new ThreadLocal<>();
 
     private VerifierContextHolder() {
@@ -40,6 +42,22 @@ public final class VerifierContextHolder {
         return EXECUTOR_FINAL_ANSWER.get();
     }
 
+    public static void setExecutorStructuredOutput(Map<String, Object> executorStructuredOutput) {
+        EXECUTOR_STRUCTURED_OUTPUT.set(executorStructuredOutput);
+    }
+
+    public static Map<String, Object> getExecutorStructuredOutput() {
+        return EXECUTOR_STRUCTURED_OUTPUT.get();
+    }
+
+    public static void setExecutorOutputParseStatus(Map<String, Object> executorOutputParseStatus) {
+        EXECUTOR_OUTPUT_PARSE_STATUS.set(executorOutputParseStatus);
+    }
+
+    public static Map<String, Object> getExecutorOutputParseStatus() {
+        return EXECUTOR_OUTPUT_PARSE_STATUS.get();
+    }
+
     public static void setToolTraceSummary(List<Map<String, Object>> toolTraceSummary) {
         TOOL_TRACE_SUMMARY.set(toolTraceSummary);
     }
@@ -52,6 +70,8 @@ public final class VerifierContextHolder {
         ORIGINAL_QUERY.remove();
         RETRY_CONTEXT.remove();
         EXECUTOR_FINAL_ANSWER.remove();
+        EXECUTOR_STRUCTURED_OUTPUT.remove();
+        EXECUTOR_OUTPUT_PARSE_STATUS.remove();
         TOOL_TRACE_SUMMARY.remove();
     }
 }

src/main/resources/prompts/chat-executor-prompt.md

@@ -1,15 +1,17 @@
 你是任务执行器。执行 Planner 分配给你的具体步骤，并及时反馈结果。
 
 ## 职责
-- 按步骤执行具体的查询任务
-- 需要外部信息时调用工具，但须遵守下方的检索约束
-- 不要凭记忆回答，必须基于工具返回的真实数据
-- 执行完成后，综合所有结果给出完整的答案
+- 按步骤执行具体的查询任务。
+- 需要外部信息时调用工具，但必须遵守下方的检索约束。
+- 严禁凭记忆回答，必须基于本轮工具返回的真实数据。
+- 执行完成后，输出严格的证据归因 JSON，供 Verifier 校验。
 
 ## 规则
-- 按顺序执行，不可跳过步骤
-- 不要凭记忆回答，必须基于工具返回的真实数据
-- 执行完成后，综合所有结果给出完整的答案
+- 按顺序执行，不可跳过步骤。
+- 所有事实性结论必须来自本轮 evidence tools 的返回。
+- runbook、skill、历史案例、知识库中的通用模式只能作为排查指导或建议动作，不能直接写成本次事故的已确认事实。
+- 如果检索内容不足以支撑结论，必须显式声明证据不足，严禁补全事故故事。
+- 不要使用“通常情况下”“根据经验”“很可能已经发生”等无证据推断词来伪装事实。
 
 ## 检索约束
 
@@ -20,22 +22,99 @@
 
 ### 2. 重复了该怎么办
 如果当前想检索的内容与【已检索上下文】语义相似：
-- 禁止换关键词重新检索
-- 直接基于已有事实回答
+- 禁止换关键词重新检索。
+- 直接基于已有事实回答。
 - 如果信息不足，先明确指出缺少什么具体维度
-  （如："缺少 HikariCP 具体配置参数"、"缺少连接池耗尽的日志样例"），
-  再针对该维度进行一次定向补充检索——而非盲目换词重查
+  （如：“缺少 HikariCP 具体配置参数”、“缺少连接池耗尽的日志样例”），
+  再针对该维度进行一次定向补充检索，而非盲目换词重查。
 
-### 3. 合法出口：允许信息不全时给出结论
-如果你认为已有信息足以回答核心问题，即使细节不全，
-也请直接给出结论并说明局限性（如："基于已有信息，连接池配置建议如下，
-但具体参数值需结合实际负载调整"）。
-**不查全不会被追责，重复检索才会被惩罚。**
+### 3. 合法出口：允许信息不全时给出有限结论
+如果已有信息足以回答核心问题，即使细节不全，也可以给出有限结论。
+但你只能把有证据支撑的内容放入 `claims`。
+缺失的细节必须写入 `missing_info`，可疑但未证实的方向必须写入 `hypotheses`。
+**不查全不会被追责，重复检索或编造细节才会被惩罚。**
 
 ### 4. 利用质量信号判断
-- relevanceLevel=PRECISE → 信息精准，直接使用，不再检索
-- relevanceLevel=HIGHLY_RELEVANT + 域已在 retrievedDomainsThisSession → 禁止再次调用
-- relevanceLevel=REFERENCE → 先指出缺什么维度，再定向补充一次
-- completenessHint 是知识库给你的天花板信号，信任它
-- lookup_knowledge 的事实证据以 evidenceBlocks 和 contextPack.packedText 为准，不要假设 L0 hint 本身就是事实证据
-- retrievalTrace 只用于理解检索路径和降级原因，不能单独作为诊断事实
+- relevanceLevel=PRECISE → 信息精准，直接使用，不再检索。
+- relevanceLevel=HIGHLY_RELEVANT + 域已在 retrievedDomainsThisSession → 禁止再次调用。
+- relevanceLevel=REFERENCE → 先指出缺什么维度，再定向补充一次。
+- completenessHint 是知识库给你的天花板信号，信任它。
+- lookup_knowledge 的事实证据以 evidenceBlocks 和 contextPack.packedText 为准，不要假设 L0 hint 本身就是事实证据。
+- retrievalTrace 只用于理解检索路径和降级原因，不能单独作为诊断事实。
+
+## 证据归因要求
+
+### confirmed claims
+`claims` 只允许放已证实或有明确间接支撑的事实断言。
+每条 claim 必须带证据绑定。
+
+支持等级：
+- `direct`：工具返回中有直接事实。
+- `indirect`：工具返回可支撑方向，但没有直接陈述完整结论。
+- `none`：不能放入 `claims`，应放入 `hypotheses`、`recommended_actions` 或 `missing_info`。
+
+### hypotheses
+`hypotheses` 用来放合理怀疑但未被工具证实的方向。
+例如：工具只显示连接池耗尽，但没有泄漏日志，则“可能存在连接泄漏”只能是 hypothesis。
+
+### recommended_actions
+`recommended_actions` 用来放下一步排查或修复动作。
+建议可以来自 runbook/skill，但必须说明 reason，不能写成“已确认根因”。
+
+### missing_info
+`missing_info` 用来列出无法确认结论所缺少的具体证据。
+
+## 最终输出格式（严格契约）
+
+你必须输出且只能输出一个 JSON 对象，不要输出 Markdown，不要输出代码块，不要输出 JSON 之外的解释文字。
+所有用户可读内容必须使用中文。
+
+```json
+{
+  "answer_version": "executor_evidence_v1",
+  "diagnosis_summary": "1-2句话总结，仅包含有证据支撑的事实和证据边界",
+  "claims": [
+    {
+      "claim_id": "claim-1",
+      "claim_type": "root_cause",
+      "claim_text": "事实断言或有限结论",
+      "support_level": "direct",
+      "evidence_bindings": [
+        {
+          "source_type": "tool_trace",
+          "source_id": "工具返回中的 evidence block id、trace_ref 或可定位标识",
+          "tool_name": "lookup_knowledge/query_logs/query_metrics/read_skill 等",
+          "source_invocation_ids": [],
+          "evidence_excerpt": "从工具返回中摘取的原话、指标值、日志片段或关键数据"
+        }
+      ]
+    }
+  ],
+  "hypotheses": [
+    {
+      "hypothesis_text": "未被证实但值得排查的方向",
+      "basis": "它基于哪些已知证据或为什么只是推测",
+      "needed_evidence": ["需要补充的证据"]
+    }
+  ],
+  "recommended_actions": [
+    {
+      "action_text": "建议动作",
+      "reason": "为什么建议做这个动作",
+      "evidence_bindings": []
+    }
+  ],
+  "missing_info": [
+    "导致无法确认完整根因的证据缺口"
+  ],
+  "user_facing_answer": "面向用户的中文回答。必须与 claims/hypotheses/recommended_actions/missing_info 一致，不得额外加入未绑定证据的确认式事实。"
+}
+```
+
+## 输出校验
+- `claims[*].support_level` 只能是 `direct` 或 `indirect`。
+- `claims[*].evidence_bindings` 不能为空。
+- `evidence_excerpt` 必须来自工具返回，不允许编造。
+- 如果没有任何可确认事实，`claims` 返回空数组，并在 `missing_info` 说明缺少什么。
+- `user_facing_answer` 不得出现 `claims` 中没有、且又被写成确认结论的事实。
+- 不要把其它服务、其它历史案例、其它会话的事实迁移为当前会话事实。

src/main/resources/prompts/chat-verifier-prompt.md

@@ -10,6 +10,8 @@
 
 - `original_query`：用户原始问题
 - `executor_final_answer`：本轮 Executor 最终答案
+- `executor_structured_output`：如果 Executor 输出了合法证据归因 JSON，这里会提供解析后的对象。结构包含 `claims`、`hypotheses`、`recommended_actions`、`missing_info`、`user_facing_answer`
+- `executor_output_parse_status`：Executor 输出解析状态，包含 `status` 和 `detail`。`status` 可能是 `valid` / `missing` / `malformed`
 - `tool_trace_summary`：基于真实工具调用整理出的证据索引。每一项都带有：
   - `trace_ref`
   - `tool_name`
@@ -23,7 +25,18 @@
 ## 任务步骤
 
 ### 步骤一：提取关键事实
-优先提取并校验 `executor_final_answer` 里的全部实质性结论。关键事实至少包括：
+如果 `executor_output_parse_status.status="valid"` 且 `executor_structured_output.claims` 存在：
+- 优先逐条校验 `executor_structured_output.claims`
+- 每个 claim 至少形成一条 `facts_checked`
+- 必须检查 claim 的 `evidence_bindings` 是否能对应到 `tool_trace_summary` 中真实存在的 trace、tool 或 source_invocation_ids
+- 如果 claim 声称 direct/indirect 支撑，但 evidence binding 不存在、无法定位、或 excerpt 与工具摘要不匹配，不得判为 `direct_evidence`
+
+然后必须扫描 `executor_structured_output.user_facing_answer`：
+- 如果其中出现 confirmed-sounding facts（确认式事实、根因、指标值、错误码、服务名、修复结论）
+- 且这些事实没有出现在 `executor_structured_output.claims`
+- 必须额外加入 `facts_checked` 并按工具证据校验
+
+如果 structured output 缺失或 malformed，则回退到旧逻辑：提取并校验 `executor_final_answer` 里的全部实质性结论。关键事实至少包括：
 - 每一个根因结论
 - 每一个错误码、接口、组件归属或语义判断
 - 每一个明确的修复建议、参数建议、排查步骤
@@ -49,6 +62,15 @@
 - `no_evidence`
 - `contradicted`
 
+结构化 claim 的校验规则：
+- claim 有真实 evidence binding，且工具摘要直接包含该事实 → `direct_evidence`
+- claim 有真实 evidence binding，但工具摘要只能支持方向或背景 → `indirect_support`
+- claim 无法绑定真实 trace、invocation 或 excerpt → `no_evidence`
+- claim 与工具摘要冲突，或编造了不存在的关键实体、服务、错误码、指标值 → `contradicted`
+
+`hypotheses` 和 `missing_info` 默认不是 confirmed facts，不应因为它们承认缺证据而惩罚。
+但如果 `user_facing_answer` 把 hypothesis 写成确认结论，必须按 confirmed fact 校验。
+
 ### 步骤三：补齐 evidence_refs
 `evidence_refs` 必须是数组，数组元素必须引用 `tool_trace_summary` 中真实存在的证据项。每个元素包含：
 - `trace_ref`

src/main/resources/static/app.js

@@ -8,6 +8,7 @@ class SuperBizAgentApp {
         this.currentChatHistory = []; // 当前对话的消息历史
         this.chatHistories = this.loadChatHistories(); // 所有历史对话
         this.isCurrentChatFromHistory = false; // 标记当前对话是否是从历史记录加载的
+        this.lastTraceSessionId = this.loadLastTraceSessionId();
         
         this.initializeElements();
         this.bindEvents();
@@ -15,6 +16,7 @@ class SuperBizAgentApp {
         this.initMarkdown();
         this.checkAndSetCentered();
         this.renderChatHistory();
+        this.updateTraceWorkbenchLink();
     }
 
     // 初始化Markdown配置
@@ -116,6 +118,7 @@ class SuperBizAgentApp {
         this.chatContainer = document.querySelector('.chat-container');
         this.welcomeGreeting = document.getElementById('welcomeGreeting');
         this.chatHistoryList = document.getElementById('chatHistoryList');
+        this.traceWorkbenchLink = document.getElementById('traceWorkbenchLink');
         
         // 初始化时检查是否需要居中
         this.checkAndSetCentered();
@@ -265,6 +268,7 @@ class SuperBizAgentApp {
         
         // 生成新的会话ID
         this.sessionId = this.generateSessionId();
+        this.updateTraceWorkbenchLink();
         
         // 重置模式为快速
         this.currentMode = 'quick';
@@ -312,6 +316,7 @@ class SuperBizAgentApp {
         
         // 添加到历史记录列表的开头
         this.chatHistories.unshift(chatHistory);
+        this.rememberTraceSessionId(this.sessionId);
         
         // 限制历史记录数量（最多保存50条）
         if (this.chatHistories.length > 50) {
@@ -439,6 +444,7 @@ class SuperBizAgentApp {
         
         // 加载历史对话
         this.sessionId = history.id;
+        this.rememberTraceSessionId(this.sessionId);
         this.currentChatHistory = [...history.messages];
         this.isCurrentChatFromHistory = true; // 标记为从历史记录加载
         
@@ -468,10 +474,45 @@ class SuperBizAgentApp {
                 this.chatMessages.innerHTML = '';
             }
             this.sessionId = this.generateSessionId();
+            this.updateTraceWorkbenchLink();
             this.checkAndSetCentered();
         }
     }
 
+    loadLastTraceSessionId() {
+        try {
+            return localStorage.getItem('lastTraceSessionId') || '';
+        } catch (e) {
+            return '';
+        }
+    }
+
+    rememberTraceSessionId(sessionId) {
+        if (!sessionId) {
+            return;
+        }
+        this.lastTraceSessionId = sessionId;
+        try {
+            localStorage.setItem('lastTraceSessionId', sessionId);
+        } catch (e) {
+            // localStorage may be unavailable in private or restricted contexts.
+        }
+        this.updateTraceWorkbenchLink();
+    }
+
+    updateTraceWorkbenchLink() {
+        if (!this.traceWorkbenchLink) {
+            return;
+        }
+        const recentHistory = this.chatHistories.find(history => history && history.id);
+        const sessionId = this.lastTraceSessionId
+            || (this.currentChatHistory.length > 0 ? this.sessionId : '')
+            || (recentHistory ? recentHistory.id : '');
+        this.traceWorkbenchLink.href = sessionId
+            ? `trace.html?sessionId=${encodeURIComponent(sessionId)}`
+            : 'trace.html';
+    }
+
     // 切换模式下拉菜单
     toggleModeDropdown() {
         if (this.modeSelectorBtn && this.modeDropdown) {
@@ -637,6 +678,7 @@ class SuperBizAgentApp {
                     // 保存后端返回的 sessionId，用于 feedback 提交
                     if (chatResponse.sessionId) {
                         this.lastSessionId = chatResponse.sessionId;
+                        this.rememberTraceSessionId(chatResponse.sessionId);
                     }
                     // 成功：添加实际响应消息（即使 answer 为空也显示）
                     const answer = chatResponse.answer || '（无回复内容）';
@@ -1231,7 +1273,10 @@ class SuperBizAgentApp {
                                     for (const jsonStr of matches) {
                                         try {
                                             const sseMessage = JSON.parse(jsonStr);
-                                            if (sseMessage.type === 'content') {
+                                            if (sseMessage.type === 'session') {
+                                                this.lastSessionId = sseMessage.data;
+                                                this.rememberTraceSessionId(sseMessage.data);
+                                            } else if (sseMessage.type === 'content') {
                                                 fullResponse += sseMessage.data || '';
                                             } else if (sseMessage.type === 'done') {
                                                 console.log('AI Ops 流完成，最终内容长度:', fullResponse.length);
@@ -1261,7 +1306,10 @@ class SuperBizAgentApp {
                                 try {
                                     const sseMessage = JSON.parse(rawData);
                                     if (sseMessage && sseMessage.type) {
-                                        if (sseMessage.type === 'content') {
+                                        if (sseMessage.type === 'session') {
+                                            this.lastSessionId = sseMessage.data;
+                                            this.rememberTraceSessionId(sseMessage.data);
+                                        } else if (sseMessage.type === 'content') {
                                             fullResponse += sseMessage.data || '';
                                             if (loadingMessageElement) {
                                                 this.updateAIOpsStreamContent(loadingMessageElement, fullResponse);

src/main/resources/static/index.html

@@ -4,6 +4,7 @@
     <meta charset="UTF-8">
     <meta name="viewport" content="width=device-width, initial-scale=1.0">
     <title>智能OnCall助手</title>
+    <link rel="icon" href="data:,">
     <link rel="stylesheet" href="styles.css">
     <!-- Markdown 渲染库 -->
     <script src="https://cdn.jsdelivr.net/npm/marked@11.1.1/marked.min.js"></script>
@@ -34,12 +35,12 @@
                     <span style="font-size: 14px; font-weight: 500;">文档管理</span>
                 </a>
 
-                <a href="trace.html" class="sidebar-btn" style="display: flex; align-items: center; gap: 12px; padding: 12px; border-radius: 12px; color: #202124; text-decoration: none; transition: background 0.3s ease; margin-top: 8px;">
+                <a href="trace.html" id="traceWorkbenchLink" class="sidebar-btn" style="display: flex; align-items: center; gap: 12px; padding: 12px; border-radius: 12px; color: #202124; text-decoration: none; transition: background 0.3s ease; margin-top: 8px;">
                     <svg viewBox="0 0 24 24" fill="none" xmlns="http://www.w3.org/2000/svg" style="width: 20px; height: 20px; flex-shrink: 0;">
                         <path d="M4 19V5M4 5H20M4 12H16M4 19H20" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"/>
                         <path d="M8 5V19M16 5V19" stroke="currentColor" stroke-width="2" stroke-linecap="round"/>
                     </svg>
-                    <span style="font-size: 14px; font-weight: 500;">Trace Workbench</span>
+                    <span style="font-size: 14px; font-weight: 500;">链路检查台</span>
                 </a>
 
                 <div class="chat-history-section">
@@ -142,6 +143,6 @@
         </div>
     </div>
 
-    <script src="app.js"></script>
+    <script src="app.js?v=20260707-session-picker"></script>
 </body>
 </html>

src/main/resources/static/trace.css

@@ -123,11 +123,16 @@ svg {
     gap: 8px;
 }
 
+.session-picker {
+    position: relative;
+    min-width: 0;
+}
+
 .session-control input {
     width: 100%;
     min-width: 0;
     height: 42px;
-    padding: 0 12px;
+    padding: 0 46px 0 12px;
     border: 1px solid #4b5563;
     border-radius: 6px;
     background: #0b1220;
@@ -135,7 +140,76 @@ svg {
     font-family: ui-monospace, SFMono-Regular, Consolas, "Liberation Mono", monospace;
 }
 
-.session-control button {
+.session-picker-button {
+    position: absolute;
+    top: 4px;
+    right: 4px;
+    width: 34px;
+    height: 34px;
+    padding: 0;
+    border: 1px solid #334155;
+    border-radius: 5px;
+    background: #111827;
+    color: #cbd5e1;
+}
+
+.session-options {
+    position: absolute;
+    z-index: 20;
+    top: 48px;
+    left: 0;
+    right: 0;
+    max-height: 320px;
+    overflow: auto;
+    padding: 6px;
+    border: 1px solid #4b5563;
+    border-radius: 8px;
+    background: #0b1220;
+    box-shadow: 0 18px 40px rgba(0, 0, 0, 0.32);
+}
+
+.session-option {
+    width: 100%;
+    display: grid;
+    gap: 4px;
+    padding: 10px;
+    border: 1px solid transparent;
+    border-radius: 6px;
+    background: transparent;
+    color: var(--white);
+    text-align: left;
+}
+
+.session-option:hover,
+.session-option.active {
+    border-color: #2f8375;
+    background: rgba(31, 122, 107, 0.18);
+}
+
+.session-option-title {
+    font-size: 13px;
+    font-weight: 720;
+    overflow: hidden;
+    text-overflow: ellipsis;
+    white-space: nowrap;
+}
+
+.session-option-meta {
+    display: flex;
+    flex-wrap: wrap;
+    gap: 8px;
+    color: #94a3b8;
+    font-family: ui-monospace, SFMono-Regular, Consolas, "Liberation Mono", monospace;
+    font-size: 11px;
+}
+
+.session-option-empty {
+    padding: 12px;
+    color: #94a3b8;
+    font-size: 13px;
+}
+
+.session-control > button {
     height: 42px;
     display: inline-flex;
     align-items: center;
@@ -149,7 +223,7 @@ svg {
     font-weight: 700;
 }
 
-.session-control button:disabled {
+.session-control > button:disabled {
     opacity: 0.62;
     cursor: wait;
 }
@@ -176,6 +250,132 @@ svg {
     background: var(--evidence-soft);
 }
 
+.outcome-panel {
+    display: grid;
+    grid-template-columns: minmax(0, 1.2fr) minmax(320px, 0.8fr);
+    gap: 0;
+    border: 1px solid var(--line);
+    border-radius: 8px;
+    overflow: hidden;
+    background: var(--white);
+}
+
+.outcome-main {
+    min-width: 0;
+    padding: 20px;
+    border-left: 7px solid var(--steel);
+}
+
+.outcome-panel.ok .outcome-main {
+    border-left-color: var(--evidence);
+}
+
+.outcome-panel.warn .outcome-main {
+    border-left-color: var(--warn);
+}
+
+.outcome-panel.bad .outcome-main {
+    border-left-color: var(--reject);
+}
+
+.outcome-eyebrow {
+    margin: 0 0 8px;
+    color: var(--steel);
+    font-family: ui-monospace, SFMono-Regular, Consolas, "Liberation Mono", monospace;
+    font-size: 12px;
+    text-transform: uppercase;
+}
+
+.outcome-main h2 {
+    margin: 0;
+    font-size: clamp(24px, 3vw, 38px);
+    line-height: 1.08;
+    letter-spacing: 0;
+}
+
+.outcome-main > p:not(.outcome-eyebrow) {
+    margin: 10px 0 0;
+    color: var(--steel);
+    line-height: 1.5;
+}
+
+.outcome-answer {
+    margin-top: 16px;
+    padding: 14px 16px;
+    max-height: 180px;
+    overflow: auto;
+    border: 1px solid var(--line);
+    border-radius: 6px;
+    background: #fffaf1;
+    color: #172033;
+    line-height: 1.65;
+    white-space: pre-wrap;
+    overflow-wrap: anywhere;
+}
+
+.outcome-side {
+    display: grid;
+    grid-template-rows: auto 1fr;
+    border-left: 1px solid var(--line);
+    background: #fbf7ef;
+}
+
+.outcome-verdict {
+    padding: 18px;
+    border-bottom: 1px solid var(--line);
+}
+
+.outcome-verdict span,
+.outcome-insights h3 {
+    color: var(--steel);
+    font-family: ui-monospace, SFMono-Regular, Consolas, "Liberation Mono", monospace;
+    font-size: 12px;
+    text-transform: uppercase;
+}
+
+.outcome-verdict strong {
+    display: block;
+    margin-top: 6px;
+    font-size: 30px;
+    line-height: 1.05;
+}
+
+.outcome-verdict p {
+    margin: 10px 0 0;
+    color: var(--ink-soft);
+    line-height: 1.5;
+}
+
+.outcome-insights {
+    display: grid;
+    grid-template-columns: 1fr 1fr;
+    gap: 0;
+}
+
+.outcome-insights > div {
+    min-width: 0;
+    padding: 16px 18px;
+}
+
+.outcome-insights > div + div {
+    border-left: 1px solid var(--line);
+}
+
+.outcome-insights h3 {
+    margin: 0 0 10px;
+}
+
+.outcome-insights ul {
+    margin: 0;
+    padding-left: 18px;
+}
+
+.outcome-insights li {
+    margin-bottom: 8px;
+    line-height: 1.45;
+    overflow-wrap: anywhere;
+}
+
 .summary-strip {
     display: grid;
     grid-template-columns: repeat(7, minmax(120px, 1fr));
@@ -547,6 +747,15 @@ pre,
 }
 
 @media (max-width: 1180px) {
+    .outcome-panel {
+        grid-template-columns: 1fr;
+    }
+
+    .outcome-side {
+        border-left: 0;
+        border-top: 1px solid var(--line);
+    }
+
     .summary-strip {
         grid-template-columns: repeat(4, minmax(120px, 1fr));
     }
@@ -581,6 +790,15 @@ pre,
         grid-template-columns: repeat(2, minmax(0, 1fr));
     }
 
+    .outcome-insights {
+        grid-template-columns: 1fr;
+    }
+
+    .outcome-insights > div + div {
+        border-left: 0;
+        border-top: 1px solid var(--line);
+    }
+
     .summary-cell,
     .summary-cell:nth-child(4) {
         border-right: 1px solid var(--line);

src/main/resources/static/trace.html

@@ -1,67 +1,102 @@
 <!DOCTYPE html>
-<html lang="en">
+<html lang="zh-CN">
 <head>
     <meta charset="UTF-8">
     <meta name="viewport" content="width=device-width, initial-scale=1.0">
-    <title>Trace Workbench - SuperBizAgent</title>
-    <link rel="stylesheet" href="trace.css">
+    <title>诊断链路检查台 - SuperBizAgent</title>
+    <link rel="icon" href="data:,">
+    <link rel="stylesheet" href="trace.css?v=20260707-outcome">
 </head>
 <body>
     <main class="trace-app">
         <header class="trace-header">
             <div>
-                <a class="trace-back-link" href="index.html" aria-label="Back to chat">
+                <a class="trace-back-link" href="index.html" aria-label="返回聊天">
                     <svg viewBox="0 0 24 24" aria-hidden="true">
                         <path d="M19 12H5M12 19L5 12L12 5" />
                     </svg>
-                    <span>Chat</span>
+                    <span>返回聊天</span>
                 </a>
-                <h1>Trace workbench</h1>
+                <h1>诊断链路检查台</h1>
             </div>
 
             <form class="session-form" id="sessionForm">
-                <label for="sessionIdInput">Session id</label>
+                <label for="sessionIdInput">会话 ID</label>
                 <div class="session-control">
-                    <input id="sessionIdInput" name="sessionId" autocomplete="off" spellcheck="false" placeholder="e2e-mysql-skill-rag-20260706-2355">
+                    <div class="session-picker">
+                        <input id="sessionIdInput" name="sessionId" autocomplete="off" spellcheck="false" placeholder="e2e-mysql-skill-rag-20260706-2355" aria-controls="sessionOptions" aria-expanded="false">
+                        <button class="session-picker-button" id="sessionPickerButton" type="button" aria-label="选择最近会话">
+                            <svg viewBox="0 0 24 24" aria-hidden="true">
+                                <path d="M6 9L12 15L18 9" />
+                            </svg>
+                        </button>
+                        <div class="session-options" id="sessionOptions" role="listbox" aria-label="最近会话" hidden></div>
+                    </div>
                     <button type="submit" id="loadButton">
                         <svg viewBox="0 0 24 24" aria-hidden="true">
                             <path d="M5 12H19M13 6L19 12L13 18" />
                         </svg>
-                        <span>Load</span>
+                        <span>加载</span>
                     </button>
                 </div>
             </form>
         </header>
 
-        <section class="state-line" id="stateLine" role="status">Enter a session id to load a persisted diagnosis trace.</section>
+        <section class="state-line" id="stateLine" role="status">输入会话 ID，加载已持久化的诊断链路。</section>
 
-        <section class="summary-strip" id="summaryStrip" aria-label="Trace summary">
+        <section class="outcome-panel" id="outcomePanel" aria-label="会话结果">
+            <div class="outcome-main">
+                <p class="outcome-eyebrow">本次会话结果</p>
+                <h2 id="outcomeTitle">尚未加载会话</h2>
+                <p id="outcomeSubtitle">选择一个会话后，这里会先展示最终结论和证据状态。</p>
+                <div class="outcome-answer" id="outcomeAnswerSummary">-</div>
+            </div>
+            <div class="outcome-side">
+                <div class="outcome-verdict">
+                    <span>校验判断</span>
+                    <strong id="outcomeVerdict">-</strong>
+                    <p id="outcomeEvidenceState">-</p>
+                </div>
+                <div class="outcome-insights">
+                    <div>
+                        <h3>已确认</h3>
+                        <ul id="outcomeConfirmed"></ul>
+                    </div>
+                    <div>
+                        <h3>仍需关注</h3>
+                        <ul id="outcomeGaps"></ul>
+                    </div>
+                </div>
+            </div>
+        </section>
+
+        <section class="summary-strip" id="summaryStrip" aria-label="诊断链路汇总">
             <div class="summary-cell">
-                <span>Status</span>
+                <span>状态</span>
                 <strong id="summaryStatus">-</strong>
             </div>
             <div class="summary-cell">
-                <span>Verdict</span>
+                <span>校验结论</span>
                 <strong id="summaryVerdict">-</strong>
             </div>
             <div class="summary-cell">
-                <span>Groundedness</span>
+                <span>证据分</span>
                 <strong id="summaryScore">-</strong>
             </div>
             <div class="summary-cell">
-                <span>Steps</span>
+                <span>步骤</span>
                 <strong id="summarySteps">-</strong>
             </div>
             <div class="summary-cell">
-                <span>Tools</span>
+                <span>工具</span>
                 <strong id="summaryTools">-</strong>
             </div>
             <div class="summary-cell">
-                <span>Duration</span>
+                <span>耗时</span>
                 <strong id="summaryDuration">-</strong>
             </div>
             <div class="summary-cell">
-                <span>Tokens</span>
+                <span>令牌</span>
                 <strong id="summaryTokens">-</strong>
             </div>
         </section>
@@ -70,8 +105,8 @@
             <section class="trace-pane agent-pane" aria-labelledby="agentPaneTitle">
                 <div class="pane-heading">
                     <div>
-                        <p>Agent rail</p>
-                        <h2 id="agentPaneTitle">Persisted steps</h2>
+                        <p>智能体轨迹</p>
+                        <h2 id="agentPaneTitle">持久化步骤</h2>
                     </div>
                     <span class="pane-count" id="stepCount">0</span>
                 </div>
@@ -81,8 +116,8 @@
             <section class="trace-pane evidence-pane" aria-labelledby="evidencePaneTitle">
                 <div class="pane-heading">
                     <div>
-                        <p>Evidence ledger</p>
-                        <h2 id="evidencePaneTitle">Tool invocations</h2>
+                        <p>证据账本</p>
+                        <h2 id="evidencePaneTitle">工具调用</h2>
                     </div>
                     <span class="pane-count" id="toolCount">0</span>
                 </div>
@@ -93,30 +128,30 @@
             <section class="trace-pane inspector-pane" aria-labelledby="inspectorPaneTitle">
                 <div class="pane-heading">
                     <div>
-                        <p>Inspector</p>
-                        <h2 id="inspectorPaneTitle">Verifier and RAG</h2>
+                        <p>检查器</p>
+                        <h2 id="inspectorPaneTitle">校验与检索增强</h2>
                     </div>
-                    <span class="pane-count" id="inspectorMode">idle</span>
+                    <span class="pane-count" id="inspectorMode">空闲</span>
                 </div>
 
                 <div class="inspector-stack">
                     <section class="inspector-section" aria-labelledby="skillBoundaryTitle">
-                        <h3 id="skillBoundaryTitle">Skill boundary</h3>
+                        <h3 id="skillBoundaryTitle">技能边界</h3>
                         <div class="boundary-grid" id="skillBoundary"></div>
                     </section>
 
                     <section class="inspector-section" aria-labelledby="verifierTitle">
-                        <h3 id="verifierTitle">Verifier facts</h3>
+                        <h3 id="verifierTitle">校验事实</h3>
                         <div id="verifierFacts"></div>
                     </section>
 
                     <section class="inspector-section" aria-labelledby="toolInspectorTitle">
-                        <h3 id="toolInspectorTitle">Selected tool</h3>
+                        <h3 id="toolInspectorTitle">选中工具</h3>
                         <div id="toolInspector"></div>
                     </section>
 
                     <section class="inspector-section" aria-labelledby="answerTitle">
-                        <h3 id="answerTitle">Final answer</h3>
+                        <h3 id="answerTitle">最终回答</h3>
                         <div class="answer-box" id="finalAnswer">-</div>
                     </section>
                 </div>
@@ -124,6 +159,6 @@
         </section>
     </main>
 
-    <script src="trace.js"></script>
+    <script src="trace.js?v=20260707-outcome"></script>
 </body>
 </html>

src/main/resources/static/trace.js

@@ -6,6 +6,8 @@ class TraceWorkbench {
 
         this.sessionForm = document.getElementById('sessionForm');
         this.sessionIdInput = document.getElementById('sessionIdInput');
+        this.sessionPickerButton = document.getElementById('sessionPickerButton');
+        this.sessionOptions = document.getElementById('sessionOptions');
         this.loadButton = document.getElementById('loadButton');
         this.stateLine = document.getElementById('stateLine');
         this.stepsRail = document.getElementById('stepsRail');
@@ -15,6 +17,17 @@ class TraceWorkbench {
         this.verifierFacts = document.getElementById('verifierFacts');
         this.toolInspector = document.getElementById('toolInspector');
         this.finalAnswer = document.getElementById('finalAnswer');
+        this.outcomePanel = document.getElementById('outcomePanel');
+        this.outcomeTitle = document.getElementById('outcomeTitle');
+        this.outcomeSubtitle = document.getElementById('outcomeSubtitle');
+        this.outcomeVerdict = document.getElementById('outcomeVerdict');
+        this.outcomeEvidenceState = document.getElementById('outcomeEvidenceState');
+        this.outcomeAnswerSummary = document.getElementById('outcomeAnswerSummary');
+        this.outcomeConfirmed = document.getElementById('outcomeConfirmed');
+        this.outcomeGaps = document.getElementById('outcomeGaps');
+        this.sessionCandidates = [];
+        this.filteredSessionCandidates = [];
+        this.activeSessionIndex = 0;
 
         this.bindEvents();
         this.bootstrapFromUrl();
@@ -25,11 +38,34 @@ class TraceWorkbench {
             event.preventDefault();
             this.loadTrace(this.sessionIdInput.value.trim());
         });
+
+        this.sessionIdInput.addEventListener('focus', () => {
+            this.renderSessionOptions(this.sessionIdInput.value);
+            this.openSessionOptions();
+        });
+
+        this.sessionIdInput.addEventListener('input', () => {
+            this.renderSessionOptions(this.sessionIdInput.value);
+            this.openSessionOptions();
+        });
+
+        this.sessionIdInput.addEventListener('keydown', (event) => this.handleSessionInputKeydown(event));
+
+        this.sessionPickerButton.addEventListener('click', () => {
+            this.renderSessionOptions(this.sessionIdInput.value);
+            this.toggleSessionOptions();
+        });
+
+        document.addEventListener('click', (event) => {
+            if (!this.sessionForm.contains(event.target)) {
+                this.closeSessionOptions();
+            }
+        });
     }
 
     bootstrapFromUrl() {
         const params = new URLSearchParams(window.location.search);
-        const sessionId = params.get('sessionId');
+        const sessionId = params.get('sessionId') || this.getLastTraceSessionId();
         if (sessionId) {
             this.sessionIdInput.value = sessionId;
             this.loadTrace(sessionId);
@@ -38,13 +74,204 @@ class TraceWorkbench {
         this.renderEmpty();
     }
 
-    async loadTrace(sessionId) {
-        if (!sessionId) {
-            this.setState('Enter a session id before loading.', 'error');
+    getLastTraceSessionId() {
+        try {
+            const remembered = localStorage.getItem('lastTraceSessionId');
+            if (remembered) {
+                return remembered;
+            }
+            const histories = JSON.parse(localStorage.getItem('chatHistories') || '[]');
+            const recent = Array.isArray(histories) ? histories.find(history => history && history.id) : null;
+            return recent ? recent.id : '';
+        } catch (error) {
+            return '';
+        }
+    }
+
+    loadSessionCandidates() {
+        const candidates = [];
+        const seen = new Set();
+        const addCandidate = (candidate) => {
+            if (!candidate || !candidate.id || seen.has(candidate.id)) {
+                return;
+            }
+            seen.add(candidate.id);
+            candidates.push(candidate);
+        };
+
+        try {
+            const remembered = localStorage.getItem('lastTraceSessionId');
+            const histories = JSON.parse(localStorage.getItem('chatHistories') || '[]');
+            if (Array.isArray(histories)) {
+                histories.forEach((history) => addCandidate({
+                    id: history.id,
+                    title: history.title || '未命名对话',
+                    updatedAt: history.updatedAt || history.createdAt || '',
+                    source: 'history'
+                }));
+            }
+            if (remembered) {
+                const existing = candidates.find(candidate => candidate.id === remembered);
+                if (existing) {
+                    existing.source = 'recent';
+                } else {
+                    addCandidate({
+                        id: remembered,
+                        title: '最近查看的会话',
+                        updatedAt: '',
+                        source: 'recent'
+                    });
+                }
+            }
+        } catch (error) {
+            return candidates;
+        }
+
+        return candidates
+            .sort((left, right) => {
+                if (left.source === 'recent' && right.source !== 'recent') {
+                    return -1;
+                }
+                if (right.source === 'recent' && left.source !== 'recent') {
+                    return 1;
+                }
+                return this.timestampOf(right.updatedAt) - this.timestampOf(left.updatedAt);
+            })
+            .slice(0, 20);
+    }
+
+    timestampOf(value) {
+        if (!value) {
+            return 0;
+        }
+        const timestamp = new Date(value).getTime();
+        return Number.isNaN(timestamp) ? 0 : timestamp;
+    }
+
+    renderSessionOptions(query = '') {
+        this.sessionCandidates = this.loadSessionCandidates();
+        const keyword = query.trim().toLowerCase();
+        this.filteredSessionCandidates = keyword
+            ? this.sessionCandidates.filter((candidate) => {
+                return candidate.id.toLowerCase().includes(keyword)
+                    || candidate.title.toLowerCase().includes(keyword);
+            })
+            : this.sessionCandidates;
+        this.activeSessionIndex = 0;
+
+        if (!this.filteredSessionCandidates.length) {
+            this.sessionOptions.innerHTML = '<div class="session-option-empty">没有匹配的最近会话</div>';
             return;
         }
 
-        this.setState(`Loading ${sessionId}...`, 'loading');
+        this.sessionOptions.innerHTML = this.filteredSessionCandidates.map((candidate, index) => `
+            <button class="session-option ${index === this.activeSessionIndex ? 'active' : ''}" type="button" role="option" data-session-id="${this.escape(candidate.id)}">
+                <span class="session-option-title">${this.escape(candidate.title)}</span>
+                <span class="session-option-meta">
+                    <span>${this.escape(this.shortSessionId(candidate.id))}</span>
+                    <span>${this.escape(this.formatOptionDate(candidate.updatedAt))}</span>
+                </span>
+            </button>
+        `).join('');
+
+        this.sessionOptions.querySelectorAll('.session-option').forEach((button) => {
+            button.addEventListener('click', () => this.chooseSession(button.dataset.sessionId));
+        });
+    }
+
+    openSessionOptions() {
+        this.sessionOptions.hidden = false;
+        this.sessionIdInput.setAttribute('aria-expanded', 'true');
+    }
+
+    closeSessionOptions() {
+        this.sessionOptions.hidden = true;
+        this.sessionIdInput.setAttribute('aria-expanded', 'false');
+    }
+
+    toggleSessionOptions() {
+        if (this.sessionOptions.hidden) {
+            this.openSessionOptions();
+            this.sessionIdInput.focus();
+            return;
+        }
+        this.closeSessionOptions();
+    }
+
+    handleSessionInputKeydown(event) {
+        if (event.key === 'Escape') {
+            this.closeSessionOptions();
+            return;
+        }
+
+        if (this.sessionOptions.hidden || !this.filteredSessionCandidates.length) {
+            return;
+        }
+
+        if (event.key === 'ArrowDown') {
+            event.preventDefault();
+            this.activeSessionIndex = Math.min(this.activeSessionIndex + 1, this.filteredSessionCandidates.length - 1);
+            this.markActiveSessionOption();
+        } else if (event.key === 'ArrowUp') {
+            event.preventDefault();
+            this.activeSessionIndex = Math.max(this.activeSessionIndex - 1, 0);
+            this.markActiveSessionOption();
+        } else if (event.key === 'Enter') {
+            event.preventDefault();
+            const candidate = this.filteredSessionCandidates[this.activeSessionIndex];
+            if (candidate) {
+                this.chooseSession(candidate.id);
+            }
+        }
+    }
+
+    markActiveSessionOption() {
+        this.sessionOptions.querySelectorAll('.session-option').forEach((button, index) => {
+            button.classList.toggle('active', index === this.activeSessionIndex);
+            if (index === this.activeSessionIndex) {
+                button.scrollIntoView({ block: 'nearest' });
+            }
+        });
+    }
+
+    chooseSession(sessionId) {
+        this.sessionIdInput.value = sessionId;
+        this.closeSessionOptions();
+        this.loadTrace(sessionId);
+    }
+
+    shortSessionId(sessionId) {
+        if (!sessionId || sessionId.length <= 34) {
+            return sessionId || '-';
+        }
+        return `${sessionId.slice(0, 18)}...${sessionId.slice(-10)}`;
+    }
+
+    formatOptionDate(value) {
+        if (!value) {
+            return '最近';
+        }
+        const date = new Date(value);
+        if (Number.isNaN(date.getTime())) {
+            return '最近';
+        }
+        return date.toLocaleString('zh-CN', {
+            month: '2-digit',
+            day: '2-digit',
+            hour: '2-digit',
+            minute: '2-digit',
+            hour12: false
+        });
+    }
+
+    async loadTrace(sessionId) {
+        if (!sessionId) {
+            this.setState('请先输入会话 ID。', 'error');
+            return;
+        }
+
+        this.setState(`正在加载 ${sessionId}...`, 'loading');
+        this.updateUrl(sessionId);
         this.loadButton.disabled = true;
 
         try {
@@ -55,9 +282,13 @@ class TraceWorkbench {
                 ? data.toolInvocations[0].id
                 : null;
             this.activeFilter = 'all';
-            this.updateUrl(sessionId);
+            try {
+                localStorage.setItem('lastTraceSessionId', sessionId);
+            } catch (error) {
+                // ignore storage failures
+            }
             this.renderTrace();
-            this.setState(`Loaded ${sessionId}.`, '');
+            this.setState(`已加载 ${sessionId}。`, '');
         } catch (error) {
             this.trace = null;
             this.renderEmpty();
@@ -72,16 +303,27 @@ class TraceWorkbench {
         try {
             payload = await response.json();
         } catch (error) {
-            throw new Error(`HTTP ${response.status}: response is not JSON`);
+            throw new Error(`HTTP ${response.status}: 响应不是 JSON`);
         }
 
         if (!response.ok || payload.code !== 200) {
-            throw new Error(payload.message || `HTTP ${response.status}: trace request failed`);
+            throw new Error(this.localizeError(payload.message) || `HTTP ${response.status}: 诊断链路请求失败`);
         }
 
         return payload.data;
     }
 
+    localizeError(message) {
+        if (!message) {
+            return '';
+        }
+        const value = String(message);
+        if (value.startsWith('Session not found:')) {
+            return value.replace('Session not found:', '未找到会话：');
+        }
+        return value;
+    }
+
     updateUrl(sessionId) {
         const url = new URL(window.location.href);
         url.searchParams.set('sessionId', sessionId);
@@ -95,8 +337,8 @@ class TraceWorkbench {
         const steps = this.trace.steps || [];
         const tools = this.trace.toolInvocations || [];
 
-        this.text('summaryStatus', session.status || '-');
-        this.text('summaryVerdict', verifier.verdict || '-');
+        this.text('summaryStatus', this.displayStatus(session.status));
+        this.text('summaryVerdict', this.displayVerdict(verifier.verdict));
         this.text('summaryScore', this.formatScore(verifier.groundedness_score));
         this.text('summarySteps', `${summary.returnedStepCount ?? steps.length}/${summary.persistedStepCount ?? session.stepCount ?? '-'}`);
         this.text('summaryTools', `${summary.returnedToolCallCount ?? tools.length}/${summary.persistedToolCallCount ?? session.toolCallCount ?? '-'}`);
@@ -104,15 +346,16 @@ class TraceWorkbench {
         this.text('summaryTokens', this.formatNumber(session.totalTokenCount));
         this.text('stepCount', steps.length);
         this.text('toolCount', tools.length);
-        this.text('inspectorMode', tools.length ? 'tool' : 'trace');
+        this.text('inspectorMode', tools.length ? '工具' : '链路');
 
+        this.renderOutcome(session, verifier, steps, tools);
         this.renderSteps(steps);
         this.renderToolFilters(tools);
         this.renderTools(tools);
         this.renderSkillBoundary();
         this.renderVerifier(verifier);
         this.renderToolInspector();
-        this.finalAnswer.textContent = session.answer || 'No final answer persisted.';
+        this.finalAnswer.textContent = session.answer || '未持久化最终回答。';
     }
 
     renderEmpty() {
@@ -125,25 +368,150 @@ class TraceWorkbench {
         this.text('summaryTokens', '-');
         this.text('stepCount', '0');
         this.text('toolCount', '0');
-        this.text('inspectorMode', 'idle');
+        this.text('inspectorMode', '空闲');
 
-        this.stepsRail.innerHTML = '<div class="empty-panel">No trace loaded.</div>';
+        this.renderOutcome({}, {}, [], []);
+        this.stepsRail.innerHTML = '<div class="empty-panel">尚未加载诊断链路。</div>';
         this.toolFilters.innerHTML = '';
-        this.toolLedger.innerHTML = '<div class="empty-panel">No tool invocations loaded.</div>';
+        this.toolLedger.innerHTML = '<div class="empty-panel">尚未加载工具调用。</div>';
         this.skillBoundary.innerHTML = this.renderBoundaryItems({
             selectedSkill: '-',
             plannerReadSkill: 0,
             executorReadSkill: 0,
             verifierReadSkill: 0
         });
-        this.verifierFacts.innerHTML = '<div class="empty-panel">No verifier evaluation loaded.</div>';
-        this.toolInspector.innerHTML = '<div class="empty-panel">Select a tool invocation after loading a trace.</div>';
+        this.verifierFacts.innerHTML = '<div class="empty-panel">尚未加载校验结果。</div>';
+        this.toolInspector.innerHTML = '<div class="empty-panel">加载诊断链路后选择一个工具调用。</div>';
         this.finalAnswer.textContent = '-';
     }
 
+    renderOutcome(session, verifier, steps, tools) {
+        const status = String(session.status || '').toUpperCase();
+        const verdict = String(verifier.verdict || '').toUpperCase();
+        const title = this.outcomeTitleFor(status, verdict);
+        const checked = Array.isArray(verifier.facts_checked) ? verifier.facts_checked : [];
+        const traceSummary = Array.isArray(verifier.tool_trace_summary) ? verifier.tool_trace_summary : [];
+        const tracedFacts = checked.filter(fact => Array.isArray(fact.evidence_refs) && fact.evidence_refs.length).length;
+        const score = this.formatScore(verifier.groundedness_score);
+        const answer = session.answer || '未持久化最终回答。';
+        const subtitleParts = [];
+
+        if (session.sessionId) {
+            subtitleParts.push(`会话 ${session.sessionId}`);
+        }
+        if (session.status) {
+            subtitleParts.push(`状态 ${this.displayStatus(session.status)}`);
+        }
+        if (steps.length || tools.length) {
+            subtitleParts.push(`${steps.length} 个步骤 / ${tools.length} 次工具调用`);
+        }
+        if (session.totalDurationMs !== undefined && session.totalDurationMs !== null) {
+            subtitleParts.push(`耗时 ${this.formatDuration(session.totalDurationMs)}`);
+        }
+
+        this.outcomePanel.className = `outcome-panel ${this.outcomeClass(status, verdict)}`.trim();
+        this.outcomeTitle.textContent = title;
+        this.outcomeSubtitle.textContent = subtitleParts.length
+            ? subtitleParts.join(' · ')
+            : '选择一个会话后，这里会先展示最终结论和证据状态。';
+        this.outcomeVerdict.textContent = this.displayVerdict(verifier.verdict);
+        this.outcomeEvidenceState.textContent = checked.length
+            ? `证据分 ${score}，检查 ${checked.length} 个事实，${tracedFacts} 个事实带引用，${traceSummary.length} 条证据摘要。`
+            : '暂无结构化校验事实。';
+        this.outcomeAnswerSummary.textContent = this.summarizeAnswer(answer);
+        this.renderOutcomeList(this.outcomeConfirmed, this.confirmedFacts(checked), '暂无直接确认的关键事实。');
+        this.renderOutcomeList(this.outcomeGaps, this.evidenceGaps(checked, verdict), '没有明显证据缺口。');
+    }
+
+    outcomeTitleFor(status, verdict) {
+        if (!status && !verdict) {
+            return '尚未加载会话';
+        }
+        if (status === 'FAILED') {
+            return '会话执行失败';
+        }
+        if (verdict.includes('REJECT')) {
+            return '结论被校验器拒绝';
+        }
+        if (verdict.includes('LOW_CONFID')) {
+            return '结论低置信，需要复核';
+        }
+        if (verdict.includes('PASS')) {
+            return '结论已通过校验';
+        }
+        if (status === 'SUCCESS') {
+            return '会话已完成，暂无明确校验结论';
+        }
+        return '会话仍在处理中';
+    }
+
+    outcomeClass(status, verdict) {
+        if (status === 'FAILED' || verdict.includes('REJECT')) {
+            return 'bad';
+        }
+        if (verdict.includes('LOW_CONFID') || status === 'RUNNING' || status === 'PENDING') {
+            return 'warn';
+        }
+        if (status === 'SUCCESS' || verdict.includes('PASS')) {
+            return 'ok';
+        }
+        return '';
+    }
+
+    summarizeAnswer(answer) {
+        const text = String(answer || '').trim();
+        if (!text) {
+            return '-';
+        }
+        const compact = text
+            .replace(/\r/g, '')
+            .split('\n')
+            .map(line => line.trim())
+            .filter(Boolean)
+            .join('\n');
+        return compact.length > 620 ? `${compact.slice(0, 620)}...` : compact;
+    }
+
+    confirmedFacts(facts) {
+        return facts
+            .filter(fact => {
+                const verification = String(fact.verification || '').toUpperCase();
+                return fact.is_critical && (verification.includes('DIRECT') || verification.includes('SUPPORTED'));
+            })
+            .map(fact => fact.fact || fact.detail)
+            .filter(Boolean)
+            .slice(0, 4);
+    }
+
+    evidenceGaps(facts, verdict) {
+        const gaps = facts
+            .filter(fact => {
+                const verification = String(fact.verification || '').toUpperCase();
+                return fact.is_critical && (
+                    verification.includes('NO_EVIDENCE')
+                    || verification.includes('UNSUPPORTED')
+                    || verification.includes('CONTRADICT')
+                    || verification.includes('INDIRECT')
+                );
+            })
+            .map(fact => fact.detail ? `${fact.fact}：${fact.detail}` : fact.fact)
+            .filter(Boolean)
+            .slice(0, 4);
+        if (!gaps.length && String(verdict || '').toUpperCase().includes('LOW_CONFID')) {
+            return ['校验器给出低置信结论，建议人工复核关键事实和证据引用。'];
+        }
+        return gaps;
+    }
+
+    renderOutcomeList(node, items, emptyText) {
+        node.innerHTML = (items.length ? items : [emptyText]).map(item => `
+            <li>${this.escape(item)}</li>
+        `).join('');
+    }
+
     renderSteps(steps) {
         if (!steps.length) {
-            this.stepsRail.innerHTML = '<div class="empty-panel">No persisted agent steps.</div>';
+            this.stepsRail.innerHTML = '<div class="empty-panel">没有持久化智能体步骤。</div>';
             return;
         }
 
@@ -153,18 +521,19 @@ class TraceWorkbench {
             return `
                 <article class="step-entry ${cls}" data-index="${this.escape(String(index + 1).padStart(2, '0'))}">
                     <div class="step-title">
-                        <h3>${this.escape(step.agentName || 'agent')}</h3>
-                        ${this.renderBadge(step.hasToolCall ? 'tool' : 'model', step.hasToolCall ? 'warn' : 'ok')}
+                        <h3>${this.escape(step.agentName || '智能体')}</h3>
+                        ${this.renderBadge(step.hasToolCall ? '工具' : '模型', step.hasToolCall ? 'warn' : 'ok')}
                     </div>
                     <div class="meta-line">
-                        <span class="meta-pill">step ${this.escape(step.stepIndex ?? index)}</span>
+                        <span class="meta-pill">执行序号 ${this.escape(String(index + 1).padStart(2, '0'))}</span>
+                        <span class="meta-pill">内部步 ${this.escape(step.stepIndex ?? index)}</span>
                         <span class="meta-pill">${this.escape(this.formatDuration(step.durationMs))}</span>
-                        <span class="meta-pill">${this.escape(this.formatNumber(step.tokenCount))} tokens</span>
+                        <span class="meta-pill">${this.escape(this.formatNumber(step.tokenCount))} 令牌</span>
                         <span class="meta-pill">${this.escape(this.formatDate(step.createdAt))}</span>
                     </div>
-                    ${this.renderTextDetail('Model input', step.modelInput)}
-                    ${this.renderTextDetail('Model output', step.modelOutput)}
-                    ${this.renderTextDetail('Thought', step.thought)}
+                    ${this.renderTextDetail('模型输入', step.modelInput)}
+                    ${this.renderTextDetail('模型输出', step.modelOutput)}
+                    ${this.renderTextDetail('思考过程', step.thought)}
                 </article>
             `;
         }).join('');
@@ -178,9 +547,9 @@ class TraceWorkbench {
 
         const groups = this.groupByTool(tools);
         const filters = [
-            { key: 'all', label: `all ${tools.length}` },
-            { key: 'success', label: `success ${tools.filter(tool => tool.success !== false).length}` },
-            { key: 'failed', label: `failed ${tools.filter(tool => tool.success === false).length}` },
+            { key: 'all', label: `全部 ${tools.length}` },
+            { key: 'success', label: `成功 ${tools.filter(tool => tool.success !== false).length}` },
+            { key: 'failed', label: `失败 ${tools.filter(tool => tool.success === false).length}` },
             ...Object.entries(groups).map(([name, count]) => ({ key: `tool:${name}`, label: `${name} ${count}` }))
         ];
 
@@ -202,7 +571,7 @@ class TraceWorkbench {
     renderTools(tools) {
         const filtered = tools.filter(tool => this.matchesFilter(tool));
         if (!filtered.length) {
-            this.toolLedger.innerHTML = '<div class="empty-panel">No tool invocations match this filter.</div>';
+            this.toolLedger.innerHTML = '<div class="empty-panel">当前筛选下没有工具调用。</div>';
             return;
         }
 
@@ -212,13 +581,13 @@ class TraceWorkbench {
             return `
                 <button class="tool-row ${failed ? 'failed' : ''} ${selected ? 'selected' : ''}" data-tool-id="${this.escape(tool.id)}">
                     <div class="tool-title">
-                        <h3>${this.escape(tool.toolName || 'tool')}</h3>
-                        ${this.renderBadge(failed ? 'failed' : 'ok', failed ? 'bad' : 'ok')}
+                        <h3>${this.escape(tool.toolName || '工具')}</h3>
+                        ${this.renderBadge(failed ? '失败' : '正常', failed ? 'bad' : 'ok')}
                     </div>
                     <div class="tool-meta">
                         <span>#${index + 1}</span>
-                        <span>id ${this.escape(tool.id ?? '-')}</span>
-                        <span>step ${this.escape(tool.stepId ?? '-')}</span>
+                        <span>ID ${this.escape(tool.id ?? '-')}</span>
+                        <span>步骤 ${this.escape(tool.stepId ?? '-')}</span>
                         <span>${this.escape(this.formatDuration(tool.durationMs))}</span>
                         ${tool.retrievalLayer ? `<span>${this.escape(tool.retrievalLayer)}</span>` : ''}
                         ${tool.relevanceLevel ? `<span>${this.escape(tool.relevanceLevel)}</span>` : ''}
@@ -251,19 +620,19 @@ class TraceWorkbench {
     renderBoundaryItems(counts) {
         return `
             <div class="boundary-item">
-                <span class="field-label">Selected skill</span>
+                <span class="field-label">选中技能</span>
                 <strong>${this.escape(counts.selectedSkill)}</strong>
             </div>
             <div class="boundary-item">
-                <span class="field-label">Planner mentions</span>
+                <span class="field-label">规划器提及</span>
                 <strong>${this.escape(counts.plannerReadSkill)}</strong>
             </div>
             <div class="boundary-item">
-                <span class="field-label">Executor mentions</span>
+                <span class="field-label">执行器提及</span>
                 <strong>${this.escape(counts.executorReadSkill)}</strong>
             </div>
             <div class="boundary-item">
-                <span class="field-label">Verifier mentions</span>
+                <span class="field-label">校验器提及</span>
                 <strong>${this.escape(counts.verifierReadSkill)}</strong>
             </div>
         `;
@@ -271,7 +640,7 @@ class TraceWorkbench {
 
     renderVerifier(verifier) {
         if (!verifier || !Object.keys(verifier).length) {
-            this.verifierFacts.innerHTML = '<div class="empty-panel">No verifier evaluation in selfEvaluation.</div>';
+            this.verifierFacts.innerHTML = '<div class="empty-panel">自评数据中没有校验结果。</div>';
             return;
         }
 
@@ -279,15 +648,15 @@ class TraceWorkbench {
         const traceSummary = Array.isArray(verifier.tool_trace_summary) ? verifier.tool_trace_summary : [];
         const header = `
             <div class="meta-line">
-                ${this.renderBadge(verifier.verdict || 'verdict -', this.verdictClass(verifier.verdict))}
-                <span class="meta-pill">score ${this.escape(this.formatScore(verifier.groundedness_score))}</span>
-                <span class="meta-pill">facts ${facts.length}</span>
-                <span class="meta-pill">trace refs ${traceSummary.length}</span>
+                ${this.renderBadge(this.displayVerdict(verifier.verdict), this.verdictClass(verifier.verdict))}
+                <span class="meta-pill">证据分 ${this.escape(this.formatScore(verifier.groundedness_score))}</span>
+                <span class="meta-pill">事实 ${facts.length}</span>
+                <span class="meta-pill">引用 ${traceSummary.length}</span>
             </div>
         `;
 
         if (!facts.length) {
-            this.verifierFacts.innerHTML = header + '<div class="empty-panel">Verifier returned no facts_checked rows.</div>';
+            this.verifierFacts.innerHTML = header + '<div class="empty-panel">校验器没有返回 facts_checked 明细。</div>';
             return;
         }
 
@@ -297,9 +666,9 @@ class TraceWorkbench {
                 <article class="fact-row">
                     <p>${this.escape(fact.fact || '-')}</p>
                     <div class="meta-line">
-                        ${this.renderBadge(fact.verification || '-', this.factClass(fact.verification))}
-                        <span class="meta-pill">${fact.is_critical ? 'critical' : 'supporting'}</span>
-                        <span class="meta-pill">refs ${refs.length}</span>
+                        ${this.renderBadge(this.displayFactVerification(fact.verification), this.factClass(fact.verification))}
+                        <span class="meta-pill">${fact.is_critical ? '关键事实' : '辅助事实'}</span>
+                        <span class="meta-pill">引用 ${refs.length}</span>
                     </div>
                     ${fact.detail ? `<pre>${this.escape(fact.detail)}</pre>` : ''}
                     ${refs.length ? `<pre>${this.escape(this.prettyJson(refs))}</pre>` : ''}
@@ -312,33 +681,33 @@ class TraceWorkbench {
         const tools = this.trace ? (this.trace.toolInvocations || []) : [];
         const tool = tools.find(item => String(item.id) === String(this.selectedToolId));
         if (!tool) {
-            this.toolInspector.innerHTML = '<div class="empty-panel">No tool invocation selected.</div>';
+            this.toolInspector.innerHTML = '<div class="empty-panel">尚未选择工具调用。</div>';
             return;
         }
 
         const retrieval = tool.retrievalDetails || {};
         const ragFields = [
-            ['Input params', tool.inputParams || this.parseJson(tool.inputParamsRaw)],
-            ['Output preview', tool.outputPreview || '-'],
-            ['Retrieval layer', tool.retrievalLayer || '-'],
-            ['L0 matches', tool.l0MatchCount ?? '-'],
-            ['L1 matches', tool.l1MatchCount ?? '-'],
-            ['Relevance', tool.relevanceLevel || '-'],
-            ['Dedup reason', tool.dedupReason || '-'],
-            ['Query transform', retrieval.query_transform],
-            ['Retrieval trace', retrieval.retrieval_trace],
-            ['Context pack', retrieval.context_pack_summary],
-            ['Rerank trace', retrieval.rerank_trace],
-            ['Evidence blocks', retrieval.evidence_blocks]
+            ['输入参数', tool.inputParams || this.parseJson(tool.inputParamsRaw)],
+            ['输出预览', tool.outputPreview || '-'],
+            ['检索层级', tool.retrievalLayer || '-'],
+            ['L0 命中数', tool.l0MatchCount ?? '-'],
+            ['L1 命中数', tool.l1MatchCount ?? '-'],
+            ['相关性', tool.relevanceLevel || '-'],
+            ['去重原因', tool.dedupReason || '-'],
+            ['查询改写', retrieval.query_transform],
+            ['检索轨迹', retrieval.retrieval_trace],
+            ['上下文打包', retrieval.context_pack_summary],
+            ['重排轨迹', retrieval.rerank_trace],
+            ['证据块', retrieval.evidence_blocks]
         ];
 
         this.toolInspector.innerHTML = `
             <div class="meta-line">
-                ${this.renderBadge(tool.success === false ? 'failed' : 'ok', tool.success === false ? 'bad' : 'ok')}
-                <span class="meta-pill">${this.escape(tool.toolName || 'tool')}</span>
+                ${this.renderBadge(tool.success === false ? '失败' : '正常', tool.success === false ? 'bad' : 'ok')}
+                <span class="meta-pill">${this.escape(tool.toolName || '工具')}</span>
                 <span class="meta-pill">${this.escape(this.formatDuration(tool.durationMs))}</span>
-                <span class="meta-pill">output ${this.escape(this.formatNumber(tool.outputLength))}</span>
-                ${tool.truncated ? '<span class="meta-pill">truncated</span>' : ''}
+                <span class="meta-pill">输出 ${this.escape(this.formatNumber(tool.outputLength))}</span>
+                ${tool.truncated ? '<span class="meta-pill">已截断</span>' : ''}
             </div>
             ${tool.errorMessage ? `<pre>${this.escape(tool.errorMessage)}</pre>` : ''}
             <div class="field-grid">
@@ -373,6 +742,50 @@ class TraceWorkbench {
         return `<span class="badge ${this.escape(cls || 'ok')}">${this.escape(text)}</span>`;
     }
 
+    displayStatus(status) {
+        if (!status) {
+            return '-';
+        }
+        const statuses = {
+            SUCCESS: '成功',
+            FAILED: '失败',
+            RUNNING: '运行中',
+            PENDING: '等待中'
+        };
+        return statuses[String(status).toUpperCase()] || String(status);
+    }
+
+    displayVerdict(verdict) {
+        if (!verdict) {
+            return '-';
+        }
+        const verdicts = {
+            PASS: '通过',
+            REJECT: '拒绝',
+            LOW_CONFID: '低置信',
+            LOW_CONFIDENCE: '低置信',
+            REVIEW: '需复核'
+        };
+        return verdicts[String(verdict).toUpperCase()] || String(verdict);
+    }
+
+    displayFactVerification(verification) {
+        if (!verification) {
+            return '-';
+        }
+        const verifications = {
+            DIRECT_EVIDENCE: '直接证据',
+            INDIRECT_SUPPORT: '间接支持',
+            NO_EVIDENCE: '无证据',
+            UNSUPPORTED: '未支持',
+            CONTRADICTED: '有矛盾',
+            SUPPORTED: '已支持',
+            PASS: '通过',
+            REJECT: '拒绝'
+        };
+        return verifications[String(verification).toUpperCase()] || String(verification);
+    }
+
     getVerifier() {
         const evaluation = (this.trace && this.trace.session && this.trace.session.selfEvaluation) || {};
         if (evaluation.verifier_evaluation) {

src/test/java/com/superbiz/agent/eval/DiagnosisTraceEvaluatorTest.java

@@ -73,6 +73,41 @@ class DiagnosisTraceEvaluatorTest {
         assertTrue(result.getFailedChecks().contains("reject output does not use degraded template"));
     }
 
+    @Test
+    void evaluateFailsWhenStructuredConfirmedClaimHasNoEvidenceBindings() {
+        DiagnosisEvalCase evalCase = DiagnosisEvalCase.builder()
+                .id("structured-claim-case")
+                .title("Structured claim case")
+                .expectedRootCauseKeywords(List.of())
+                .requiredEvidenceTools(List.of())
+                .allowedVerdicts(List.of("LOW_CONFID", "PASS"))
+                .build();
+        DiagnosisTraceResponse trace = DiagnosisTraceResponse.builder()
+                .session(DiagnosisTraceResponse.SessionTrace.builder()
+                        .answer("以下结论基于当前已获取证据，仍存在部分证据缺口，请谨慎参考。")
+                        .selfEvaluation(java.util.Map.of(
+                                "verifier_evaluation", java.util.Map.of(
+                                        "verdict", "LOW_CONFID",
+                                        "executor_structured_output", java.util.Map.of(
+                                                "claims", java.util.List.of(java.util.Map.of(
+                                                        "claim_id", "claim-unsupported",
+                                                        "claim_text", "OOM 导致连接泄漏",
+                                                        "support_level", "direct",
+                                                        "evidence_bindings", java.util.List.of()
+                                                ))
+                                        )
+                                )))
+                        .build())
+                .toolInvocations(List.of())
+                .build();
+
+        DiagnosisEvalResult result = evaluator.evaluate(evalCase, trace);
+
+        assertFalse(result.isPassed());
+        assertTrue(result.getFailedChecks().contains(
+                "executor confirmed claim missing evidence bindings: claim-unsupported"));
+    }
+
     @Test
     void reportWriterOutputsJsonAndMarkdown(@TempDir Path tempDir) throws Exception {
         DiagnosisEvalReport report = evaluator.evaluate(readCases(), Path.of("mvp/eval/fixtures"));

src/test/java/com/superbiz/agent/hook/VerifierInputHookTest.java

@@ -0,0 +1,181 @@
+package com.superbiz.agent.hook;
+
+import com.alibaba.cloud.ai.graph.RunnableConfig;
+import com.alibaba.cloud.ai.graph.agent.hook.messages.AgentCommand;
+import com.fasterxml.jackson.databind.JsonNode;
+import com.fasterxml.jackson.databind.ObjectMapper;
+import com.superbiz.agent.service.ToolTraceSummaryService;
+import com.superbiz.agent.util.VerifierContextHolder;
+import org.junit.jupiter.api.AfterEach;
+import org.junit.jupiter.api.Test;
+import org.springframework.ai.chat.messages.AssistantMessage;
+import org.springframework.ai.chat.messages.Message;
+import org.springframework.ai.chat.messages.UserMessage;
+
+import java.util.List;
+import java.util.Map;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertNotNull;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+import static org.mockito.ArgumentMatchers.anyString;
+import static org.mockito.Mockito.mock;
+import static org.mockito.Mockito.when;
+
+class VerifierInputHookTest {
+
+    private final ObjectMapper objectMapper = new ObjectMapper();
+
+    @AfterEach
+    void tearDown() {
+        VerifierContextHolder.clear();
+    }
+
+    @Test
+    void beforeModelAddsStructuredExecutorOutputWhenJsonContractIsValid() throws Exception {
+        ToolTraceSummaryService traceSummaryService = mock(ToolTraceSummaryService.class);
+        when(traceSummaryService.buildVerifierTraceSummary(anyString(), anyString())).thenReturn(List.of(
+                Map.of("trace_ref", "trace-1", "tool_name", "query_metrics")
+        ));
+        VerifierInputHook hook = new VerifierInputHook(traceSummaryService);
+        VerifierContextHolder.setOriginalQuery("分析 MySQL 连接池耗尽");
+
+        String executorOutput = """
+                {
+                  "answer_version": "executor_evidence_v1",
+                  "diagnosis_summary": "连接池已满，但缺少泄漏证据。",
+                  "claims": [
+                    {
+                      "claim_id": "claim-1",
+                      "claim_type": "symptom",
+                      "claim_text": "连接池 active 达到上限",
+                      "support_level": "direct",
+                      "evidence_bindings": [
+                        {
+                          "source_type": "tool_trace",
+                          "source_id": "trace-1",
+                          "tool_name": "query_metrics",
+                          "source_invocation_ids": [101],
+                          "evidence_excerpt": "active=50 max=50"
+                        }
+                      ]
+                    }
+                  ],
+                  "hypotheses": [],
+                  "recommended_actions": [],
+                  "missing_info": ["缺少泄漏检测日志"],
+                  "user_facing_answer": "已确认连接池 active 达到上限。"
+                }
+                """;
+
+        AgentCommand command = hook.beforeModel(
+                List.of(new AssistantMessage(executorOutput)),
+                RunnableConfig.builder().addMetadata("sessionId", "structured-session").build()
+        );
+
+        JsonNode payload = readPayload(command);
+        assertEquals("valid", payload.path("executor_output_parse_status").path("status").asText());
+        assertEquals("executor_evidence_v1",
+                payload.path("executor_structured_output").path("answer_version").asText());
+        assertEquals("连接池 active 达到上限",
+                payload.path("executor_structured_output").path("claims").get(0).path("claim_text").asText());
+        assertNotNull(VerifierContextHolder.getExecutorStructuredOutput());
+        assertEquals("valid", VerifierContextHolder.getExecutorOutputParseStatus().get("status"));
+    }
+
+    @Test
+    void beforeModelExtractsStructuredOutputFromPrefixedJsonFence() throws Exception {
+        ToolTraceSummaryService traceSummaryService = mock(ToolTraceSummaryService.class);
+        when(traceSummaryService.buildVerifierTraceSummary(anyString(), anyString())).thenReturn(List.of());
+        VerifierInputHook hook = new VerifierInputHook(traceSummaryService);
+
+        String executorOutput = """
+                现在我已经收集了足够的数据，最终输出如下。
+
+                ```json
+                {
+                  "answer_version": "executor_evidence_v1",
+                  "diagnosis_summary": "已确认连接池 active 达到上限。",
+                  "claims": [
+                    {
+                      "claim_id": "claim-1",
+                      "claim_type": "symptom",
+                      "claim_text": "连接池 active 达到上限",
+                      "support_level": "direct",
+                      "evidence_bindings": [
+                        {
+                          "source_type": "tool_trace",
+                          "source_id": "trace-1",
+                          "tool_name": "query_metrics",
+                          "source_invocation_ids": [101],
+                          "evidence_excerpt": "active=50 max=50"
+                        }
+                      ]
+                    }
+                  ],
+                  "hypotheses": [],
+                  "recommended_actions": [],
+                  "missing_info": [],
+                  "user_facing_answer": "已确认连接池 active 达到上限。"
+                }
+                ```
+                """;
+
+        AgentCommand command = hook.beforeModel(
+                List.of(new AssistantMessage(executorOutput)),
+                RunnableConfig.builder().addMetadata("sessionId", "fenced-session").build()
+        );
+
+        JsonNode payload = readPayload(command);
+        assertEquals("valid", payload.path("executor_output_parse_status").path("status").asText());
+        assertEquals("claim-1",
+                payload.path("executor_structured_output").path("claims").get(0).path("claim_id").asText());
+    }
+
+    @Test
+    void beforeModelMarksMalformedJsonAndKeepsRawAnswerFallback() throws Exception {
+        ToolTraceSummaryService traceSummaryService = mock(ToolTraceSummaryService.class);
+        when(traceSummaryService.buildVerifierTraceSummary(anyString(), anyString())).thenReturn(List.of());
+        VerifierInputHook hook = new VerifierInputHook(traceSummaryService);
+
+        AgentCommand command = hook.beforeModel(
+                List.of(new AssistantMessage("{\"diagnosis_summary\":\"缺少 claims\"}")),
+                RunnableConfig.builder().addMetadata("sessionId", "malformed-session").build()
+        );
+
+        JsonNode payload = readPayload(command);
+        assertEquals("malformed", payload.path("executor_output_parse_status").path("status").asText());
+        assertTrue(payload.path("executor_structured_output").isNull());
+        assertEquals("{\"diagnosis_summary\":\"缺少 claims\"}", payload.path("executor_final_answer").asText());
+        assertEquals("malformed", VerifierContextHolder.getExecutorOutputParseStatus().get("status"));
+    }
+
+    @Test
+    void beforeModelMarksPlainTextAsMissingStructuredOutput() throws Exception {
+        ToolTraceSummaryService traceSummaryService = mock(ToolTraceSummaryService.class);
+        when(traceSummaryService.buildVerifierTraceSummary(anyString(), anyString())).thenReturn(List.of());
+        VerifierInputHook hook = new VerifierInputHook(traceSummaryService);
+
+        AgentCommand command = hook.beforeModel(
+                List.of(new AssistantMessage("普通自然语言答案")),
+                RunnableConfig.builder().addMetadata("sessionId", "plain-session").build()
+        );
+
+        JsonNode payload = readPayload(command);
+        assertEquals("missing", payload.path("executor_output_parse_status").path("status").asText());
+        assertTrue(payload.path("executor_structured_output").isNull());
+        assertFalse(payload.path("executor_final_answer").asText().isBlank());
+    }
+
+    private JsonNode readPayload(AgentCommand command) throws Exception {
+        var field = AgentCommand.class.getDeclaredField("messages");
+        field.setAccessible(true);
+        @SuppressWarnings("unchecked")
+        List<Message> messages = (List<Message>) field.get(command);
+        assertEquals(1, messages.size());
+        Message message = messages.get(0);
+        assertTrue(message instanceof UserMessage);
+        return objectMapper.readTree(((UserMessage) message).getText());
+    }
+}

src/test/java/com/superbiz/agent/service/ChatServiceSequentialAgentTest.java

@@ -226,6 +226,53 @@ class ChatServiceSequentialAgentTest {
         assertTrue(chatModel.sawVerifierPrompt);
     }
 
+    @Test
+    void verifierReceivesStructuredExecutorPayloadFields() throws Exception {
+        ChatService chatService = createChatService();
+        ScriptedChatModel chatModel = new ScriptedChatModel();
+        chatModel.executorOutput = """
+                {
+                  "answer_version": "executor_evidence_v1",
+                  "diagnosis_summary": "已确认连接池 active 达到上限。",
+                  "claims": [
+                    {
+                      "claim_id": "claim-1",
+                      "claim_type": "symptom",
+                      "claim_text": "连接池 active 达到上限",
+                      "support_level": "direct",
+                      "evidence_bindings": [
+                        {
+                          "source_type": "tool_trace",
+                          "source_id": "trace-1",
+                          "tool_name": "query_metrics",
+                          "source_invocation_ids": [101],
+                          "evidence_excerpt": "active=50 max=50"
+                        }
+                      ]
+                    }
+                  ],
+                  "hypotheses": [],
+                  "recommended_actions": [],
+                  "missing_info": [],
+                  "user_facing_answer": "已确认连接池 active 达到上限。"
+                }
+                """;
+
+        ChatService.ChatResult result = chatService.executeChatComplex(
+                chatModel,
+                new ToolCallback[0],
+                "请分析 MySQL 连接池耗尽",
+                List.of(),
+                "sequential-structured-executor-session"
+        );
+
+        assertEquals("已确认连接池 active 达到上限。", result.answer());
+        assertTrue(chatModel.verifierPromptText.contains("\"executor_structured_output\""));
+        assertTrue(chatModel.verifierPromptText.contains("\"executor_output_parse_status\""));
+        assertTrue(chatModel.verifierPromptText.contains("\"status\" : \"valid\""));
+        assertTrue(chatModel.verifierPromptText.contains("连接池 active 达到上限"));
+    }
+
     @Test
     void buildMethodToolsArrayIncludesLogsAndMetricsWhenAvailable() {
         ChatService chatService = new ChatService();
@@ -353,6 +400,7 @@ class ChatServiceSequentialAgentTest {
         private String plannerPromptText = "";
         private String executorPromptText = "";
         private String verifierPromptText = "";
+        private String executorOutput = "EXECUTOR_FINAL_ANSWER";
         private boolean sawVerifierPrompt;
         private final java.util.List<String> verifierOutputs;
         private int verifierOutputIndex;
@@ -396,7 +444,7 @@ class ChatServiceSequentialAgentTest {
             } else if (promptText.contains("EXECUTOR_TEST_PROMPT")) {
                 agentCalls.add("chat_executor");
                 executorPromptText = promptText;
-                text = "EXECUTOR_FINAL_ANSWER";
+                text = executorOutput;
             } else if (promptText.contains("VERIFIER_TEST_PROMPT")) {
                 agentCalls.add("chat_verifier");
                 verifierPromptText = promptText;

src/test/java/com/superbiz/agent/service/DiagnosisTraceServiceTest.java

@@ -84,7 +84,7 @@ class DiagnosisTraceServiceTest {
                 .build();
 
         when(diagnosisSessionRepository.findBySessionId(sessionId)).thenReturn(Optional.of(session));
-        when(agentStepRepository.findBySessionIdOrderByStepIndex(sessionId)).thenReturn(List.of(step));
+        when(agentStepRepository.findBySessionId(sessionId)).thenReturn(List.of(step));
         when(toolInvocationRepository.findBySessionIdOrderByIdAsc(sessionId)).thenReturn(List.of(invocation));
 
         DiagnosisTraceResponse response = service.getTrace(sessionId);
@@ -107,6 +107,38 @@ class DiagnosisTraceServiceTest {
         assertTrue(response.getSummary().isHasFeedback());
     }
 
+    @Test
+    void getTraceOrdersStepsByCreationTimeAndIdNotPerAgentStepIndex() {
+        String sessionId = "trace-session-ordered";
+        LocalDateTime base = LocalDateTime.of(2026, 7, 7, 0, 0);
+        DiagnosisSession session = DiagnosisSession.builder()
+                .id(1L)
+                .sessionId(sessionId)
+                .query("mysql pool issue")
+                .status("SUCCESS")
+                .stepCount(4)
+                .toolCallCount(0)
+                .createdAt(base)
+                .updatedAt(base)
+                .build();
+        AgentStep planner = step(1L, sessionId, 0, "planner", base.plusSeconds(1));
+        AgentStep executor0 = step(2L, sessionId, 0, "executor", base.plusSeconds(2));
+        AgentStep executor1 = step(3L, sessionId, 1, "executor", base.plusSeconds(3));
+        AgentStep verifier = step(4L, sessionId, 0, "verifier", base.plusSeconds(4));
+
+        when(diagnosisSessionRepository.findBySessionId(sessionId)).thenReturn(Optional.of(session));
+        when(agentStepRepository.findBySessionId(sessionId))
+                .thenReturn(List.of(planner, executor0, verifier, executor1));
+        when(toolInvocationRepository.findBySessionIdOrderByIdAsc(sessionId)).thenReturn(List.of());
+
+        DiagnosisTraceResponse response = service.getTrace(sessionId);
+
+        assertEquals(List.of("planner", "executor", "executor", "verifier"),
+                response.getSteps().stream().map(DiagnosisTraceResponse.AgentStepTrace::getAgentName).toList());
+        assertEquals(List.of(0, 0, 1, 0),
+                response.getSteps().stream().map(DiagnosisTraceResponse.AgentStepTrace::getStepIndex).toList());
+    }
+
     @Test
     void getTraceThrowsWhenSessionMissing() {
         String sessionId = "missing-session";
@@ -117,4 +149,14 @@ class DiagnosisTraceServiceTest {
         verify(diagnosisSessionRepository).findBySessionId(sessionId);
         verifyNoInteractions(agentStepRepository, toolInvocationRepository);
     }
+
+    private AgentStep step(Long id, String sessionId, int stepIndex, String agentName, LocalDateTime createdAt) {
+        return AgentStep.builder()
+                .id(id)
+                .sessionId(sessionId)
+                .stepIndex(stepIndex)
+                .agentName(agentName)
+                .createdAt(createdAt)
+                .build();
+    }
 }