feat: add needs-human PR status labels and unified review marker (#2589)

.claude/skills/multi-pr-review/SKILL.md

@@ -103,7 +103,7 @@ Options:
 #### Example Summary Comment
 
 ```markdown
-## :mag: Multi-Agent Code Review
+## :mag: Dyadbot Code Review Summary
 
 Found **4** new issue(s) flagged by 3 independent reviewers.
 (2 issue(s) skipped - already commented)
@@ -133,7 +133,7 @@ Found **4** new issue(s) flagged by 3 independent reviewers.
 
 See inline comments for details.
 
-_Generated by multi-agent consensus review_
+_Generated by Dyadbot code review_
 ```
 
 ## File Structure

.claude/skills/multi-pr-review/scripts/post_comment.py

@@ -186,7 +186,7 @@ def format_summary_comment(
     low_issues = [i for i in issues if i.get('severity') == 'LOW']
 
     lines = [
-        "## :mag: Multi-Agent Code Review",
+        "## :mag: Dyadbot Code Review Summary",
         "",
     ]
 
@@ -196,7 +196,7 @@ def format_summary_comment(
             lines.append(f":white_check_mark: No new issues found. ({num_duplicates} issue(s) already commented on)")
         else:
             lines.append(":white_check_mark: No issues found by consensus review.")
-        lines.extend(["", "*Generated by multi-agent consensus review*"])
+        lines.extend(["", "*Generated by Dyadbot code review*"])
         return "\n".join(lines)
 
     total_new = len(issues)
@@ -267,7 +267,7 @@ def format_summary_comment(
         lines.append("See inline comments for details.")
         lines.append("")
 
-    lines.append("*Generated by multi-agent consensus review*")
+    lines.append("*Generated by Dyadbot code review*")
 
     return "\n".join(lines)
 

.claude/skills/swarm-pr-review/SKILL.md

@@ -215,7 +215,7 @@ Based on the confirmed issues:
 Post a summary comment on the PR using `gh pr comment`:
 
 ```markdown
-## :mag: Swarm Code Review
+## :mag: Dyadbot Code Review Summary
 
 **Verdict: [VERDICT EMOJI + TEXT]**
 
@@ -245,7 +245,7 @@ Reviewed by 3 specialized agents: Correctness Expert, Code Health Expert, UX Wiz
 </details>
 
 ---
-*Generated by swarm code review — 3 agents collaborated on this review*
+*Generated by Dyadbot code review*
 ````
 
 #### Inline Comments

.github/workflows/pr-review-responder.yml

@@ -105,22 +105,9 @@ jobs:
             // Guard: do not loop more than 3 retries (4 total runs: initial + 3 retries)
             // cc:request (0) -> cc:request:1 (1) -> cc:request:2 (2) -> cc:request:3 (3) -> blocked
             if (requestCount >= 4) {
-              console.log(`PR #${prNumber} has reached max retry count (${requestCount}), adding cc:needs-human-review`);
+              console.log(`PR #${prNumber} has reached max retry count (${requestCount}), adding needs-human:review-issue`);
 
-              // Ensure the label exists
-              try {
-                await github.rest.issues.createLabel({
-                  owner: context.repo.owner,
-                  repo: context.repo.repo,
-                  name: 'cc:needs-human-review',
-                  color: 'd93f0b',
-                  description: 'Claude Code has exhausted auto-retries; needs human review'
-                });
-              } catch (e) {
-                // Label already exists, ignore
-              }
-
-              // Remove current label, cc:pending, and add needs-human-review
+              // Remove current label, cc:pending
               await github.rest.issues.removeLabel({
                 owner: context.repo.owner,
                 repo: context.repo.repo,
@@ -135,13 +122,22 @@ jobs:
                 name: 'cc:pending'
               }).catch(() => {});
 
+              // Also add needs-human:review-issue (retries exhausted = needs human attention).
               await github.rest.issues.addLabels({
                 owner: context.repo.owner,
                 repo: context.repo.repo,
                 issue_number: prNumber,
-                labels: ['cc:needs-human-review']
+                labels: ['needs-human:review-issue']
               });
 
+              await github.rest.issues.removeLabel({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                issue_number: prNumber,
+                name: 'needs-human:final-check'
+              }).catch(() => {});
+
+              core.setOutput('pr_number', prNumber);
               core.setOutput('should_continue', 'false');
               return;
             }
@@ -248,26 +244,6 @@ jobs:
             echo "commits_pushed=false" >> $GITHUB_OUTPUT
           fi
 
-      - name: Ensure next request label exists
-        # Use always() so retry loop continues even if Claude Code fails after pushing commits
-        if: steps.pr-info.outputs.should_continue == 'true' && always() && steps.retrigger.outputs.commits_pushed == 'true'
-        uses: actions/github-script@v7
-        with:
-          script: |
-            const nextCount = parseInt('${{ steps.pr-info.outputs.request_count }}', 10) + 1;
-            const labelName = `cc:request:${nextCount}`;
-            try {
-              await github.rest.issues.createLabel({
-                owner: context.repo.owner,
-                repo: context.repo.repo,
-                name: labelName,
-                color: '1d76db',
-                description: `Claude Code auto-re-request iteration ${nextCount}`
-              });
-            } catch (e) {
-              // Label already exists, ignore
-            }
-
       - name: Update labels - re-request review (commits pushed)
         # Use always() so retry loop continues even if Claude Code fails after pushing commits
         if: steps.pr-info.outputs.should_continue == 'true' && always() && steps.retrigger.outputs.commits_pushed == 'true'
@@ -292,3 +268,36 @@ jobs:
           gh pr edit ${{ steps.pr-info.outputs.pr_number }} --repo ${{ github.repository }} --remove-label "cc:pending" --add-label "cc:failed"
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Checkout base repo for labeler script
+        # The earlier checkout (line 186) checks out the PR head repo, which for
+        # fork PRs may not contain scripts/pr-status-labeler.js. Checkout the
+        # base repo's default branch to a separate path so the script is always available.
+        if: >-
+          steps.pr-info.outputs.should_continue == 'true' &&
+          always() &&
+          steps.retrigger.outputs.commits_pushed != 'true'
+        uses: actions/checkout@v5
+        with:
+          path: __base
+          sparse-checkout: scripts/pr-status-labeler.js
+
+      - name: Apply needs-human status label
+        # Run when PR reaches a terminal state (cc:done or cc:failed) but NOT when
+        # commits were pushed (retry loop continues). Uses always() to run even if
+        # the Claude Code step failed.
+        if: >-
+          steps.pr-info.outputs.should_continue == 'true' &&
+          always() &&
+          steps.retrigger.outputs.commits_pushed != 'true'
+        uses: actions/github-script@v7
+        with:
+          script: |
+            const { run } = require('./__base/scripts/pr-status-labeler.js');
+            await run({
+              github,
+              context,
+              core,
+              prNumber: parseInt('${{ steps.pr-info.outputs.pr_number }}', 10),
+              ciConclusion: '${{ github.event.workflow_run.conclusion }}'
+            });

.github/workflows/pr-status-labeler.yml

+name: PR Status Labeler
+
+on:
+  workflow_run:
+    workflows: ["CI"]
+    types: [completed]
+
+permissions:
+  contents: read
+  pull-requests: write
+  issues: write
+
+jobs:
+  label-pr:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Find PR and check eligibility
+        id: pr-info
+        uses: actions/github-script@v7
+        with:
+          script: |
+            const run = context.payload.workflow_run;
+
+            // Find associated PR (same approach as playwright-comment.yml)
+            let prNumber;
+
+            if (run.pull_requests && run.pull_requests.length > 0) {
+              prNumber = run.pull_requests[0].number;
+            } else {
+              if (!run.head_repository) {
+                console.log('Head repository not available');
+                core.setOutput('should_continue', 'false');
+                return;
+              }
+              const { data: prs } = await github.rest.pulls.list({
+                owner: context.repo.owner,
+                repo: context.repo.repo,
+                state: 'open',
+                head: `${run.head_repository.owner.login}:${run.head_branch}`
+              });
+              if (prs.length === 0) {
+                console.log('No pull requests found for this workflow run');
+                core.setOutput('should_continue', 'false');
+                return;
+              }
+              prNumber = prs[0].number;
+            }
+
+            // Fetch full PR details to check labels
+            const { data: pr } = await github.rest.pulls.get({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              pull_number: prNumber
+            });
+
+            // Skip PRs actively managed by pr-review-responder (cc:request* or cc:pending)
+            // PRs with cc:done/cc:failed/needs-human-review ARE eligible —
+            // a human may have pushed new commits after the automation loop ended.
+            const activeLabels = pr.labels.filter(l =>
+              l.name.startsWith('cc:request') || l.name === 'cc:pending'
+            );
+
+            if (activeLabels.length > 0) {
+              console.log(`PR #${prNumber} is in the cc: retry loop (${activeLabels.map(l => l.name).join(', ')}), skipping`);
+              core.setOutput('should_continue', 'false');
+              return;
+            }
+
+            console.log(`PR #${prNumber} is eligible for status labeling`);
+            core.setOutput('should_continue', 'true');
+            core.setOutput('pr_number', prNumber);
+
+      - name: Checkout repository (default branch for trusted scripts)
+        if: steps.pr-info.outputs.should_continue == 'true'
+        uses: actions/checkout@v4
+
+      - name: Apply status label
+        if: steps.pr-info.outputs.should_continue == 'true'
+        uses: actions/github-script@v7
+        with:
+          script: |
+            const { run } = require('./scripts/pr-status-labeler.js');
+            await run({
+              github,
+              context,
+              core,
+              prNumber: parseInt('${{ steps.pr-info.outputs.pr_number }}', 10),
+              ciConclusion: '${{ github.event.workflow_run.conclusion }}'
+            });

scripts/pr-status-labeler.js

+// Shared logic for applying needs-human:* labels to PRs based on CI status and code review results.
+// Used by both pr-review-responder.yml (for cc:request PRs) and pr-status-labeler.yml (for all other PRs).
+
+const LABEL_REVIEW_ISSUE = "needs-human:review-issue";
+const LABEL_FINAL_CHECK = "needs-human:final-check";
+
+const REVIEW_MARKER = "Dyadbot Code Review Summary";
+
+// Review verdict strings — keep in sync with:
+//   Swarm verdicts: .claude/skills/swarm-pr-review/SKILL.md
+//   Multi-agent output: .claude/skills/multi-pr-review/scripts/post_comment.py
+const SWARM_VERDICT_CLEAN = "YES - Ready to merge";
+const SWARM_VERDICT_UNSURE = "NOT SURE - Potential issues";
+const SWARM_VERDICT_REJECT = "NO - Do NOT merge";
+const MULTI_AGENT_NO_ISSUES = ":white_check_mark: No issues found";
+const MULTI_AGENT_NO_NEW_ISSUES = ":white_check_mark: No new issues found";
+// Severity table regexes match "| :emoji: LEVEL | N |" rows with non-zero counts
+const HIGH_ISSUES_RE = /:red_circle:.*?\|\s*[1-9]/;
+const MEDIUM_ISSUES_RE = /:yellow_circle:.*?\|\s*[1-9]/;
+const LOW_ISSUES_RE = /:green_circle:.*?\|\s*\d/;
+
+function findLatestReviewComment(comments) {
+  for (let i = comments.length - 1; i >= 0; i--) {
+    const body = comments[i].body || "";
+    const user = comments[i].user || {};
+    if (body.includes(REVIEW_MARKER) && user.type === "Bot") {
+      return comments[i];
+    }
+  }
+  return null;
+}
+
+function isReviewClean(body) {
+  // Swarm verdict: explicit clean
+  if (body.includes(SWARM_VERDICT_CLEAN)) {
+    return true;
+  }
+
+  // Multi-agent: no issues found
+  if (
+    body.includes(MULTI_AGENT_NO_ISSUES) ||
+    body.includes(MULTI_AGENT_NO_NEW_ISSUES)
+  ) {
+    return true;
+  }
+
+  // If there are HIGH or MEDIUM severity markers with non-zero counts, review has issues.
+  // The severity table always renders rows like "| :red_circle: HIGH | 0 |" even at count 0,
+  // so we match only rows where the count is >= 1.
+  if (body.match(HIGH_ISSUES_RE) || body.match(MEDIUM_ISSUES_RE)) {
+    return false;
+  }
+
+  // Multi-agent: severity table present with only LOW issues (HIGH=0 and MEDIUM=0
+  // already passed the regex check above, so reaching here means only LOW remain)
+  if (body.match(LOW_ISSUES_RE)) {
+    return true;
+  }
+
+  // Swarm verdicts indicating issues
+  if (
+    body.includes(SWARM_VERDICT_UNSURE) ||
+    body.includes(SWARM_VERDICT_REJECT)
+  ) {
+    return false;
+  }
+
+  // No clear signal — fail-closed: flag for human review rather than
+  // silently treating an unrecognized format as clean.
+  return false;
+}
+
+async function applyLabel(github, owner, repo, prNumber, addLabel) {
+  const removeLabel =
+    addLabel === LABEL_REVIEW_ISSUE ? LABEL_FINAL_CHECK : LABEL_REVIEW_ISSUE;
+
+  // Atomically swap labels using setLabels to avoid a window where both exist
+  const { data: currentLabels } = await github.rest.issues.listLabelsOnIssue({
+    owner,
+    repo,
+    issue_number: prNumber,
+  });
+
+  const newLabelSet = new Set(currentLabels.map((label) => label.name));
+  newLabelSet.delete(removeLabel);
+  newLabelSet.add(addLabel);
+
+  await github.rest.issues.setLabels({
+    owner,
+    repo,
+    issue_number: prNumber,
+    labels: [...newLabelSet],
+  });
+}
+
+async function run({ github, context, core, prNumber, ciConclusion }) {
+  const owner = context.repo.owner;
+  const repo = context.repo.repo;
+
+  // Bail on cancelled/skipped runs — inconclusive
+  if (ciConclusion === "cancelled" || ciConclusion === "skipped") {
+    core.info(`CI conclusion is '${ciConclusion}', skipping label update`);
+    return;
+  }
+
+  const ciSuccess = ciConclusion === "success";
+
+  // Fetch all PR comments (paginated) to find the latest code review summary
+  const comments = await github.paginate(github.rest.issues.listComments, {
+    owner,
+    repo,
+    issue_number: prNumber,
+  });
+
+  const reviewComment = findLatestReviewComment(comments);
+
+  if (!reviewComment && ciSuccess) {
+    core.info("CI passed but no review comment found, skipping label update");
+    return;
+  }
+
+  if (!reviewComment && !ciSuccess) {
+    core.info(
+      "CI failed and no review comment found, adding review-issue label",
+    );
+    await applyLabel(github, owner, repo, prNumber, LABEL_REVIEW_ISSUE);
+    return;
+  }
+
+  // Check if the review is stale (posted before the latest commit)
+  const { data: pull } = await github.rest.pulls.get({
+    owner,
+    repo,
+    pull_number: prNumber,
+  });
+  const { data: headCommit } = await github.rest.repos.getCommit({
+    owner,
+    repo,
+    ref: pull.head.sha,
+  });
+
+  const commitDate = new Date(headCommit.commit.committer.date);
+  const reviewDate = new Date(reviewComment.created_at);
+
+  if (reviewDate < commitDate) {
+    core.info(
+      "Latest review is stale (posted before latest commit), adding review-issue label",
+    );
+    await applyLabel(github, owner, repo, prNumber, LABEL_REVIEW_ISSUE);
+    return;
+  }
+
+  const reviewClean = isReviewClean(reviewComment.body);
+
+  if (ciSuccess && reviewClean) {
+    core.info("CI passed and review is clean, adding final-check label");
+    await applyLabel(github, owner, repo, prNumber, LABEL_FINAL_CHECK);
+  } else {
+    core.info(
+      `CI ${ciSuccess ? "passed" : "failed"}, review ${reviewClean ? "clean" : "has issues"}, adding review-issue label`,
+    );
+    await applyLabel(github, owner, repo, prNumber, LABEL_REVIEW_ISSUE);
+  }
+}
+
+module.exports = { run };