Skip to content

Add opt-in structured review result to local review (#69) - #527

Merged
amirbena merged 2 commits into
mainfrom
feat/local-review-structured-output
Sep 24, 2026
Merged

amirbena merged 2 commits into
mainfrom
feat/local-review-structured-output

Conversation

@amirbena

@amirbena amirbena commented Sep 24, 2026 •

Copy link
Copy Markdown
Owner

Adds an opt-in machine-readable result to local-code-review; contract and semantics live in shared/policies/structured-output.md and docs/review-result/.

Fixes #69

What

  • Behavior / contracts: New option structured_review_result (default off). When set, one schema-versioned JSON result (with schema_version and reviewed head SHA) is appended after the unchanged human report. Findings, severity, coverage and the mechanical Decision are identical on and off. reviewed_head_sha is null when the reviewed target includes uncommitted changes.
  • Governance / policy: New structured-output capability; the packaged policy also carries the finding-identity minting recipe (SHA-256 computed by shell, never written by hand). A drift test pins it to the Define review output schema #67 schema and identity reference model; the option name is protected from being read as human_review_output's structured review negative phrase.
  • Packaging / portability: New shared policy in both archives (adapters local, github; only local wires it). Manifest and skill metadata regenerated; local baseline hashes re-captured.
  • Changelog: generated at release from the two lines below — never edit CHANGELOG.md (see docs/RELEASE.md)
  • Release category: Added
  • Release entry: local-code-review can append an opt-in, schema-versioned machine-readable review result via structured_review_result.

Validation

  • Relevant validation was run, or the reason it could not be run is stated.
  • Targeted policy/unit/packaging tests and package-skills.sh all pass. Last full run had a live-model benchmark test (test_blocking_verdict_corpus real-run) failing on a different case each run; not confirmed against main. Full suite not re-run after the final fixes.

Review

Merged main (incl. #70 GitHub structured output); resolved conflicts in docs/review-result/README.md and review-result-model.md to describe both Skills' emission. The two structured-output.md policies (shared/ for local, skills/github-pr-review/policies/) are separate files.

local-code-review on the working tree: Review Clean after fixing one P1 (option-name phrase collision) and one P2 (example added); one P2 open — reviewed_head_sha is null for uncommitted targets, maintainer to confirm.

🤖 Generated with Claude Code

local-code-review can now append one schema-versioned machine-readable
JSON result after its unchanged human report when the invocation sets
structured_review_result (default off). A packaged policy owns the
serialization and finding-identity minting; a drift test pins it to the
#67 schema and the identity reference model.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
@amirbena amirbena self-assigned this Sep 24, 2026
…uctured-output

# Conflicts:
#	docs/review-result/README.md
#	docs/review-result/review-result-model.md
@amirbena
amirbena merged commit ce46dba into main Sep 24, 2026
10 checks passed
@amirbena
amirbena deleted the feat/local-review-structured-output branch September 24, 2026 13:43
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Add structured output to local review

1 participant