Normalize Claude session diagnostics and snapshot evidence by pelikhan · Pull Request #67351 · github/gh-aw

pelikhan · 2026-10-10T05:32:49Z

Why

Live Claude transcripts contain duplicate provider-error envelopes and permission-denial notices presented as assistant answers. This aligns those observations with the unified agent session contract while preserving native evidence and partial traces.

Changes

Map provider errors, retries, authentication failures, and permission denials to canonical session diagnostics. Reconcile terminal denial snapshots by observed identity without inventing tool completions or discarding distinct notices.
Retain mapped streaming envelopes and finalized snapshots on canonical events instead of emitting duplicate claude.* records. Preserve refusal policy metadata, observed partials, reasoning signatures, and recovered snapshot suffixes.
Preserve structured images, documents, and redacted reasoning. Explicit tool failure and interruption signals take precedence over contradictory success fields. Unknown native extensions and genuinely unmapped controls remain available.

Includes sanitized regression coverage and Claude-specific documentation. Shared implementation, types, projection, and specification files are unchanged; shared test edits only update Claude-specific expectations.

Evidence and limitations

Inspected existing success, API failure, documentation/subagent, and dynamic-workflow artifacts. The documentation run persisted 15 denial notices as assistant answers in both canonical and unified traces; the corrected replay retains 15 canonical denials once. The failure replay preserves all 22 error observations without duplicate native envelopes.

The newer two runs contain agent-session.jsonl and aw_session.jsonl; older runs only support native-log replay. Streaming and structured-content regressions use synthetic SDK/API shapes, not live coverage claims. No workflow runs were triggered.

Validation

make fmt-cjs && make lint-cjs
PATH=/opt/homebrew/bin:$PATH make agent-report-progress: build, lint, schema freshness, and 237 impacted JavaScript tests passed; no Go changes.
Expanded Claude and dependent shared suites: 672 tests passed across 18 matched suites.
npm run typecheck and npm run schema:session:check from actions/setup/js.
Four complete downloaded transcripts passed schema validation, JSON round-trip/idempotence checks, and exact native text, reasoning, and tool-output comparisons against canonical and projected events.

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

github-actions · 2026-10-10T05:35:04Z

✅ Design Decision Gate 🏗️ completed the design decision gate check. See the comment below for the result and any generated ADR draft.

No ADR enforcement needed for PR #67351: the PR does not carry the 'implementation' label and has 0 new lines in default business logic directories (threshold: 100, no custom .design-gate.yml present).

🏗️ ADR gate enforced by Design Decision Gate 🏗️

github-actions · 2026-10-10T05:35:05Z

🧠 Matt Pocock Skills Reviewer is reviewing this pull request using Matt Pocock's engineering skills...

github-actions · 2026-10-10T05:35:06Z

✅ Ponytail Reviewer completed successfully!

Lean already. Ship.

Generated by Ponytail Reviewer for #67351

github-actions · 2026-10-10T05:35:10Z

✅ Test Quality Sentinel completed test quality analysis.

Test Quality Sentinel skipped because pre-fetch PR data was unavailable: unable to fetch test file diff

🧪 Test quality analysis by Test Quality Sentinel

github-actions · 2026-10-10T05:35:13Z

✅ PR Code Quality Reviewer completed the code quality review.

🔎 Code quality review by PR Code Quality Reviewer

Copilot

🟡 Changes recommended

Stream normalization still duplicates recognized blocks and leaves several mapped transport envelopes as separate native events.

1 open finding

Prevent duplicate emission of streamed atomic blocks · New

What changed in this PR

Normalizes Claude diagnostics and streamed evidence into the unified session model while preserving native metadata.

Changes:

Canonicalizes errors, permission denials, refusals, and structured content.
Reconciles streamed events with finalized snapshots.
Adds regression coverage and Claude documentation.

File	Description
`.changeset/patch-claude-session-audit.md`	Records the patch release changes.
`actions/setup/js/agent_session_render.test.cjs`	Updates stream projection expectations.
`actions/setup/js/claude_session.cjs`	Implements Claude session normalization.
`actions/setup/js/claude_session.test.cjs`	Updates streaming snapshot coverage.
`actions/setup/js/claude_session_audit.test.cjs`	Adds diagnostic and evidence regressions.
`actions/setup/js/claude_session_normalization.test.cjs`	Updates denial normalization coverage.
`actions/setup/js/provider_refusal.test.cjs`	Tests attached refusal snapshots.
`actions/setup/js/unified_session.test.cjs`	Tests canonical provider errors.
`docs/src/content/docs/engines/claude.md`	Documents Claude session mapping.

🧠 Review effort: Balanced

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

github-actions · 2026-10-10T05:41:54Z

Comment Memory

Peek at saved memory (pr-code-quality-reviewer)

reviewed_at: 2026-10-10T05:36:54Z
review_event: REQUEST_CHANGES
top_themes:
  - finalized snapshot evidence loss
  - late refusal snapshot loss
  - camelCase tool success misclassification
files_reviewed:
  - actions/setup/js/claude_session.cjs
  - actions/setup/js/unified_session.cjs
  - actions/setup/js/agent_session.cjs
  - actions/setup/js/provider_refusal.cjs
comment_count: 3

Note

This comment is managed by comment memory.

Expand the saved memory block to view or edit the persistent context for this thread.
Edit only the text inside the backtick fences; workflow metadata and the footer are regenerated automatically.

Learn more about comment memory

🔎 Code quality review by PR Code Quality Reviewer · copilot · gpt54 · 49.3 AIC · ⌖ 7.87 AIC · ⊞ 19.9K · ◷
_{Comment /review to run again}

github-actions

Request changes

These changes still regress Claude trace fidelity in three places: successful camelCase tool results can become "unknown", and the only finalized SDK snapshot can be lost for both streamed blocks and late refusals.

Blocking themes

isError: false is not normalized symmetrically with the new failure aliases, so some successful tool results no longer project as explicit successes.
First finalized SDK snapshots are not always retained on the canonical event once duplicate claude.assistant_snapshot records are removed.
Late refusal reclassification drops the originating finalized snapshot, which defeats the evidence-preservation goal of this change.

🔎 Code quality review by PR Code Quality Reviewer · copilot · gpt54 · 49.3 AIC · ⌖ 7.87 AIC · ⊞ 19.9K
_{Comment /review to run again}

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

done

Normalize Claude session diagnostics and snapshot evidence

31faa52

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

pelikhan marked this pull request as ready for review October 10, 2026 05:34

Copilot AI balanced review requested due to automatic review settings October 10, 2026 05:34

Copilot started reviewing on behalf of pelikhan October 10, 2026 05:34 View session

Copilot AI reviewed Oct 10, 2026

View reviewed changes

Comment thread actions/setup/js/claude_session.cjs

github-actions Bot mentioned this pull request Oct 10, 2026

[aw] No-Op Runs #65943

Open

github-actions Bot previously requested changes Oct 10, 2026

View reviewed changes

Comment thread actions/setup/js/claude_session.cjs Outdated

Comment thread actions/setup/js/claude_session.cjs

Comment thread actions/setup/js/claude_session.cjs

github-actions Bot mentioned this pull request Oct 10, 2026

[pr-review] Daily PR Code Quality Review — 38028794652 #67366

Closed

Preserve Claude atomic blocks and first snapshot evidence

476ed40

Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

github-actions Bot mentioned this pull request Oct 10, 2026

[pr-sous-chef] PR Sous Chef Report – 2026-10-10T02:09:05Z #67318

Closed

pelikhan merged commit caa19db into main Oct 10, 2026
43 checks passed

pelikhan deleted the pelikhan-claude-session-audit branch October 10, 2026 10:25

Navigation Menu

Repository navigation

Normalize Claude session diagnostics and snapshot evidence - #67351

Normalize Claude session diagnostics and snapshot evidence#67351
pelikhan merged 2 commits into
mainfrom
pelikhan-claude-session-audit

pelikhan commented Oct 10, 2026

Uh oh!

github-actions Bot commented Oct 10, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 10, 2026

Uh oh!

github-actions Bot commented Oct 10, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 10, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 10, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

github-actions Bot commented Oct 10, 2026

Uh oh!

github-actions Bot left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

pelikhan commented Oct 10, 2026

Why

Changes

Evidence and limitations

Validation

Uh oh!

github-actions Bot commented Oct 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Oct 10, 2026

Uh oh!

github-actions Bot commented Oct 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Oct 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Oct 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Uh oh!

github-actions Bot commented Oct 10, 2026

Comment Memory

Uh oh!

github-actions Bot left a comment

Choose a reason for hiding this comment

Request changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

github-actions Bot commented Oct 10, 2026 •

edited

Loading

github-actions Bot commented Oct 10, 2026 •

edited

Loading

github-actions Bot commented Oct 10, 2026 •

edited

Loading

github-actions Bot commented Oct 10, 2026 •

edited

Loading