Vp Pr Comment Resolver

Pull Request Comment Resolver

Reviewer feedback is input, not authority. Verify every claim against the current head, repository rules, code, and tests before deciding.

Workflow

1. Build a complete two-surface snapshot: all issue comments in the PR conversation, plus the pull request review surface containing all submitted reviews and all review threads with every inline review comment and reply. Exhaust pagination independently for issue comments, submitted reviews, review threads, and nested replies; retain resolved and outdated items for context; and record counts and retrieval cursors. An incomplete surface is a blocker, not evidence that feedback is absent.
2. Classify each item as fix, already handled, no fix, disagreement, uncertain, or out of scope.
3. Apply only validated fixes, verify them, and commit by coherent topic.
4. Reply with evidence through the supported mechanism. Reply directly to a review thread or PR conversation issue comment. GitHub has no reply mutation for a top-level submitted-review body, so answer it in the PR conversation with the review author and link identified.
5. Resolve handled bot review threads. Leave human review threads unresolved unless the user explicitly directs otherwise. Submitted reviews and PR conversation issue comments have no resolvable state.
6. Re-fetch feedback and report remaining risk or required user judgment.

Convergence for automated feedback

Keep bot feedback tied to the requested change, its acceptance criteria, and the evidence needed for safe delivery. Fix verified correctness or regression problems within scope. When current code or proportionate test evidence disproves a finding, explain that evidence and close the eligible bot thread instead of changing correct code. Defer speculative hardening, preference-only refactors, and unrelated improvements once the delivery criteria are satisfied.

Batch related findings into a coherent correction pass. New feedback reopens only the affected decision and verification unless it reveals a broader problem. Do not keep expanding implementation or verification merely because a bot can suggest another improvement.

Pause for user judgment when feedback reveals a severe security, data integrity, compatibility, or operational risk; requires a material product or architectural decision; or would substantially expand the agreed scope or cost. Otherwise, make the narrowest evidence-backed decision and continue the authorized workflow.

Within authorized feedback handling, explain a demonstrably incorrect or already-handled claim with current-code or test evidence without asking again. Ask when unresolved evidence prevents a safe decision, a disagreement requires a material product or architectural decision, feedback reveals a severe risk or substantial scope or cost expansion, or history rewriting or an external write is not already authorized. Human threads still require explicit direction to resolve. Read only the relevant reference:

• references/workflow.md
• references/decision-tree.md
• references/author-classification.md
• references/commit-strategy.md
• references/reply-templates.md
• references/github-cli.md

Related skills

• vp-pr-briefing when the PR needs orientation before feedback is resolved.
• vp-git for commit strategy and pull-request lifecycle decisions.
• vp-pr-review-followup when acting as the reviewer instead of the PR author.