Read every open code review comment, group them by risk / file / theme, and present a phased fix plan for the user to approve before any edits run. Use when the review backlog is large (rough cutoff ≥ 5 open comments), when the user says "triage the review", "plan the fixes", "what's the plan for the review backlog", "group these comments", or before invoking review-workflow on a big batch. Requires the code-review MCP server to be connected. Read-only — does not edit files or call mark_resolved.
How this skill is triggered — by the user, by Claude, or both
Slash command
/code-review-annotator:triage-and-planThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use this skill before `review-workflow` when the open-comment backlog is large enough that a head-first fix loop would be reckless. The goal is to surface the *shape* of the work — risk, ordering, who needs to decide what — and get the user's go-ahead before edits start.
Use this skill before review-workflow when the open-comment backlog is large enough that a head-first fix loop would be reckless. The goal is to surface the shape of the work — risk, ordering, who needs to decide what — and get the user's go-ahead before edits start.
review-workflow directly. Planning overhead exceeds the win.Fetch the full backlog:
get_review_comments() // all open
get_review_comments({ status: 'resolved' }) // optional context: what was already done
get_tool_calls({ status: 'orphan' }) // optional: any captures with missing post-snapshot
Bucket by axes — for each comment, jot down:
Group into phases — propose 2 to 4 phases ordered by risk and file affinity:
Group line-scope comments on the same file into the same phase to avoid line-number drift between edits.
Present the plan to the user as a compact table:
<id> <file>:<line> — <body excerpt>Stop and wait. Ask explicitly: "Approve the plan, or want to re-shuffle? I won't touch anything until you say so." Do not call mark_resolved, do not edit files, do not call reply_to_comment. This skill is purely planning.
On approval → hand off to review-workflow, working through the approved phases in order. If the user asks to skip a phase or take a different order, follow that.
replies from claude that were never resolved, treat it as higher risk — the previous attempt may have failed verify.viewContext.toRef is WORKTREE references live state — re-resolve sourceLines at fix time, don't trust your initial read in the planning phase.cargo test, npm test, etc.) so the user can budget time before approving.npx claudepluginhub jason-hchsieh/code-review-annotator --plugin code-review-annotatorGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.