What problem does it solve? Handling review feedback on your own pull request is chaotic: comments mix bugs, style nits, requirement changes, and CI failures, and some touch high-risk surfaces like API contracts, schemas, or prompt definitions that demand human sign-off. This Skill fetches PR reviews via the GitHub CLI, triages every comment, and produces a structured response plan so nothing is missed or silently changed. ## Core Features & Use Cases - Comment classification: Categorizes each review comment into bug, suggestion, style, architecture, requirement, test, or CI failure, with P0/P1/P2 priority levels. - Impact assessment: Evaluates effects on linked Plan/SPEC/ADR documents plus mj-agent-specific high-risk surfaces (in-source canonical skills, biz_catalog, SQL guardrail, system.md version), escalating to HITL pause when requirement, API, schema, permission, or user-visible behavior is touched. - Reply drafting: Generates per-comment reply drafts (accept, push-back with technical reasoning, clarify, defer-to-followup, CI-fix) plus a modification plan with target files, commit shape, and retest commands. - Use Case: After a reviewer requests changes on your mj-agent PR and mypy fails in CI, run this Skill to get a per-comment table, a High-risk HITL question for the API change, and ready-to-paste reply drafts. ## Quick Start Ask the assistant to process the review comments on your current mj-agent pull request and draft responses for each comment.
