[JacobStephens2]

Summary

Require runnable evidence for correctness findings on both review axes, and evidence for every decline. The shared instruction reaches Fresh, Continuation, Spec review and the Repair that reviews Foreign commits; the generated prompts page and glossary match.

Standards / Spec finding
  behaviour is wrong → run the written test or command
    fails as claimed → fix the code + keep the test
    passes and exercises the finding → decline may cite the run
  other decline → cite refuting code / plan decision / ADR
  Spec or Day shift must decide → reuse or file a needs-triage issue → link it

New decision issues carry only needs-triage and describe the finding, evidence, originating pull request and why the call is outside the agent's authority. Both reviewer briefs include one shared evidence rule, with 800 words of prose and full tests/commands outside the limit.

Closes #488

Evidence

• Before: cargo test prompts_page failed both stale-output checks (6 passed, 2 failed). cargo test --test fresh_run failed gives_the_agent_the_fresh_prompt against the old literal prompt (18 passed, 1 failed).
  After: regenerated outputs pass all 8 prompts_page tests; the updated fresh prompt expectation passes all 19 fresh_run tests.
• Validation: cargo check, cargo fmt --check, cargo clippy --all-targets -- -D warnings, and full cargo test passed (1344 tests, 0 failures).
• Review against issue-486 at e68dcee: one Standards duplication finding fixed and verified; no Spec findings. Both axes verified the follow-up changes.

Seams: the existing prompts-page golden files and the existing Fresh Run end-to-end test that observes the Session prompt sent to the Harness. The skill text was reviewed manually, as #486 specifies.

Unaddressed findings

Standards

None.

Spec

None.

Merge Danger

Door: two-way

Reverting restores the previous review instructions; no data migration.

Blast Radius: review

Factory reviewer briefs and the four Session prompt types that run code review, including the two Continuation variants.

Built with codex · gpt-6.1-sol · xhigh