bug: review-cycle artifacts can be written without required YAML frontmatter

[robertDouglass]

Observed

During mission release-320-workflow-reliability-01KQKV85, a WP06 rejection artifact existed as review-cycle-1.md but lacked YAML frontmatter. The subsequent fix-mode prompt warned that it could not read the expected structured artifact metadata and fell through to the full prompt path.

Expected

Every review-cycle-N.md artifact should be schema-valid and include required frontmatter such as verdict, cycle, mission/WP identity, and feedback metadata. If artifact generation fails validation, the rejection command should fail before committing or persisting the pointer.

Impact

Focused fix-mode becomes unreliable: implementers may not receive the review-specific rejection context, and stale/malformed artifacts can also confuse status and merge gates.

Related

This is distinct from stale rejected verdict tracking (#904): this issue is about missing or malformed review artifact metadata at write time.

[robertDouglass]

Closing as fixed by PR #969 (merged as 0a2c1bc) and released in v3.2.0a10.\n\nRelevant verification from the PR/release line:\n- review-cycle domain and artifact coverage: tests/review/test_cycle.py, tests/review/test_artifacts.py\n- rejection transition coverage: tests/integration/review/test_reject_from_in_review.py\n- fix-mode/canonical pointer coverage: tests/agent/test_workflow_review_cycle_pointer.py\n- finalized task-board routing coverage: tests/next/test_finalized_task_routing.py\n- missing retrospective record coverage: tests/cli/test_agent_retrospect_missing_record.py and tests/cli/test_agent_retrospect_synthesize.py\n- focused smoke coverage: tests/integration/test_implement_review_retrospect_smoke.py\n\nThe PR passed CI, diff coverage, and quality gate before merge; v3.2.0a10 was published to GitHub Releases and PyPI.