Warning: This build has drifted.
The coverage report for this pull request build may be inaccurate because its base commit is no longer the HEAD of its target branch.
This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.

    • Learn more: For more information on this, see Tracking coverage changes for pull request builds.
    • Fix now: For a quick fix, rebase this PR at GitHub. Your next report should be accurate.
    • Prevent going forward: To avoid this issue with future PRs, see these Recommended CI Configurations.

fix: align default review agents with the review-loop contract

the five default agent bodies described what to review but not the
contract the surrounding loop depends on, which made review iterations
churn: agents could report pre-existing repo-wide issues every pass
(stalling the zero-findings REVIEW_DONE exit), edit files themselves
(the parent loop owns fixing and committing), and had no defined output
for the clean case. the implementation agent verified goal fit without
ever being given the goal, and the documentation agent instructed the
reviewer to edit plan files mid-review.

- review-context lead-in now carries the shared reviewer contract:
  diff scope, read-only, evidence bar, and the NO ISSUES FOUND sentinel
  already used by external reviewers - applies to custom agents too
- implementation: receives {{GOAL}}/{{PLAN_FILE}}, drops the generic
  edge-case overlap with quality, adds severity for the second pass
- quality: drops the simplicity section (simplification agent owns it,
  and it produced discardable findings in the critical-only second
  pass), dedupes checklist items, adds severity
- testing: scoped to changed code, collapses self-duplicating sections,
  location field now covers missing-test findings, adds severity and a
  focus line, allows running the suite read-only
- simplification: scope + plan-awareness, requires project-wide search
  evidence for 'unused/no callers' claims, skips generated code,
  defers large-effort findings to follow-up plans
- documentation: plan-file bullets become report items instead of edit
  imperatives, checks current docs before reporting gaps, also catches
  docs made stale by the change, findings anchored to a file:line
  trigger so the parent can verify them

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>