coderabbit-triage

name	coderabbit-triage
description	Fetch all coderabbitai[bot] review threads via GraphQL, classify each (bug/nit/suggestion/opinion/wrong), emit accept/reject/discuss plan. Triggers: triage coderabbit, classify cr comments, cr plan.

coderabbit-triage

Read all CR threads → classify each → emit a structured plan the mastermind can execute.

Fetch threads (GraphQL — REST cannot return isResolved)

PR="${1:-$(gh pr view --json number -q .number)}"
OWNER=$(gh repo view --json owner -q .owner.login)
REPO=$(gh repo view --json name -q .name)

gh api graphql -F owner="$OWNER" -F repo="$REPO" -F pr="$PR" -f query='
query($owner:String!, $repo:String!, $pr:Int!) {
  repository(owner:$owner, name:$repo) {
    pullRequest(number:$pr) {
      reviewThreads(first:100) {
        nodes {
          id
          isResolved
          isOutdated
          path
          line
          comments(first:50) {
            nodes {
              id
              databaseId
              author { login }
              body
              createdAt
              url
            }
          }
        }
        pageInfo { hasNextPage endCursor }
      }
    }
  }
}' > /tmp/cr-threads.json

Pagination: if hasNextPage, re-run with -F cursor="$endCursor" and after:$cursor in the query.

Filter to CR-authored threads

jq '[.data.repository.pullRequest.reviewThreads.nodes[]
  | select(.comments.nodes[0].author.login == "coderabbitai")
  | select(.isResolved == false)]' /tmp/cr-threads.json > /tmp/cr-open.json

Classification table (apply per thread)

Class	Definition	Default action
bug	reproducible defect, principle violation, security issue	accept
nit	style/naming/comment with no behavior change	accept if cheap, reject if churn
suggestion	alternative impl, equivalent correctness	discuss
opinion	taste/style without spec backing	reject with principle cite
wrong	factually incorrect, contradicts spec or principle	reject with principle cite
outdated	thread .isOutdated == true	resolve, no reply

Heuristics

• Body matches panic!|unwrap()|expect( outside #[cfg(test)] → bug (Law 6, Law 10).
• Body suggests changing a spec without ADR → wrong (Law 2).
• Body suggests adding logging via println! → wrong (Law 11 wants structured telemetry).
• Body suggests removing #[deny(missing_docs)] → wrong (house rule).
• Body suggests broader unsafe → wrong unless // SAFETY: proven.
• Body suggests a rename with no callers affected → nit, accept.
• Body suggests an algorithm change with measurable perf delta → suggestion, discuss + bench.

Output (JSON, one object per thread)

{
  "pr": 123,
  "threads": [
    {
      "thread_id": "PRRT_kwDOA...",
      "path": "crates/renderer/src/shadow.rs",
      "line": 142,
      "class": "bug|nit|suggestion|opinion|wrong|outdated",
      "action": "accept|reject|discuss|resolve",
      "summary": "<one-line>",
      "principle_cite": "docs/architecture/01-principles.md#law-6",
      "reply_draft": "<text to post if rejecting>",
      "comment_url": "https://github.com/..."
    }
  ],
  "counts": { "accept": 4, "reject": 2, "discuss": 1, "resolve": 0 }
}

Vendor honesty

CR is wrong sometimes. Rejecting is fine and expected. Every reject MUST cite a Nexus principle by anchor (docs/architecture/01-principles.md#law-N) so future readers see the reasoning.

Hand-off

• accept → fix-from-coderabbit
• reject/discuss → coderabbit-reply
• resolve (outdated) → coderabbit-resolve

Refs

• https://docs.github.com/en/graphql/reference/objects#pullrequestreviewthread
• https://docs.github.com/en/graphql/reference/objects#pullrequestreviewcomment
• docs/architecture/01-principles.md
• .claude/skills/gh-graphql-helpers/SKILL.md