Skip to content

fix(#537): add remediation consistency rule to correctness sub-agent - #538

Open
fullsend-ai-coder[bot] wants to merge 1 commit into
mainfrom
agent/537-remediation-consistency-rule
Open

fix(#537): add remediation consistency rule to correctness sub-agent#538
fullsend-ai-coder[bot] wants to merge 1 commit into
mainfrom
agent/537-remediation-consistency-rule

Conversation

@fullsend-ai-coder

Copy link
Copy Markdown
Contributor

Summary

  • Adds a "Remediation consistency" rule to the correctness sub-agent definition (skills/pr-review/sub-agents/correctness.md)
  • Prevents the agent from suggesting remediations that contradict its own severity analysis (e.g., suggesting a fallback to a state it described as "already broken")
  • Adds guidance for companion-PR scenarios to frame remediation around merge coordination rather than broken fallbacks

Context

On PR #527, the correctness sub-agent repeatedly raised an [api-contract] finding where it downgraded severity because a prior approach was "already broken," then simultaneously suggested falling back to that broken approach. This contradiction required 8+ human dismissals. The root cause was missing self-consistency guidance between severity rationale and remediation suggestions.

Changes

The new rule is placed under the "Runtime mechanism checklist" section, after the existing failure-paths bullet. This location is appropriate because api-contract findings are in the correctness sub-agent's domain, and the runtime mechanism checklist is where it evaluates cross-component contracts.

Testing

  • Verified the rule text matches the issue's proposed change
  • Confirmed placement is after the failure-paths bullet in the Runtime mechanism checklist
  • gitlint passed on the commit message
  • Secret scan passed

Closes #537

Post-script verification

  • Branch is not main/master (agent/537-remediation-consistency-rule)
  • Secret scan passed (gitleaks — 5e45ff31e2ccc1c69f0126f2f690aaef8e9712df..HEAD)
  • PR body secret scan passed (gitleaks — no-git)
  • Pre-commit hooks passed (authoritative run on runner)
  • Tests ran inside sandbox

Add guidance to the correctness sub-agent definition preventing
it from suggesting remediations that contradict its own analysis.
When a severity rationale acknowledges a prior state is broken,
the agent must not suggest falling back to that broken state.
When a companion PR provides the missing piece, the agent should
frame remediation around merge coordination instead.

The rule is placed under the "Runtime mechanism checklist" section
in correctness.md, after the failure-paths bullet, where the
agent evaluates cross-component contracts and API findings.

Note: pre-commit could not run in sandbox (network restriction).
The post-script runs it authoritatively on the runner.

Closes #537
@fullsend-ai-coder
fullsend-ai-coder Bot requested a review from a team as a code owner July 29, 2026 21:55
@fullsend-ai-review

fullsend-ai-review Bot commented Jul 29, 2026

Copy link
Copy Markdown

🤖 Finished Review · ✅ Success · Started 9:56 PM UTC · Completed 10:07 PM UTC
Commit: 76bc5d6 · View workflow run →

@fullsend-ai-review

Copy link
Copy Markdown

Review

Findings

Medium


Labels: PR modifies review agent sub-agent definition (skills/pr-review/sub-agents/correctness.md)

@fullsend-ai-review fullsend-ai-review Bot added requires-manual-review Review requires human judgment review-agent labels Jul 29, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Correctness sub-agent should not suggest remediation that contradicts its own analysis

0 participants