From agent-loop
Scans open GitHub issues, PRs, and CI queues, then classifies items by autonomy, review need, or deferral for autonomous loop orchestration.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agent-loop:agent-loop-triageThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Scan open issues and PRs for the current repo (or a configured set of repos).
Scan open issues and PRs for the current repo (or a configured set of repos). Classify each item and return a structured queue summary suitable for the orchestrator loop's dispatch phase.
gh auth status 2>/dev/null || { echo "gh not authenticated"; exit 1; }
repo=$(gh repo view --json nameWithOwner --jq .nameWithOwner 2>/dev/null || true)
if [ -z "$repo" ]; then
url=$(git remote get-url origin 2>/dev/null || true)
repo=$(printf '%s\n' "$url" |
sed -E 's#^[email protected]:##; s#^https://github.com/##; s#\.git$##')
fi
printf '%s\n' "$repo"
git status --short --branch
git branch --show-current
git pull --ff-only
git status --short --branch
Proceed only on main with clean worktree. If dirty or non-main, report and
skip — do not branch/stash/commit without instruction.
gh issue list --repo "$repo" --state open --limit 50 \
--json number,title,author,labels,createdAt,updatedAt,url
gh pr list --repo "$repo" --state open --limit 50 \
--json number,title,author,isDraft,reviewDecision,mergeStateStatus,createdAt,updatedAt,url
For items that look actionable:
# Issue detail
gh issue view <n> --repo "$repo" \
--json number,title,author,body,comments,labels,createdAt,updatedAt,url
# PR detail
gh pr view <n> --repo "$repo" \
--json number,title,author,body,comments,files,commits,isDraft,reviewDecision,mergeStateStatus,statusCheckRollup,createdAt,updatedAt,url
# PR diff for small changes
gh pr diff <n> --repo "$repo" --patch
Classify each item into one of:
Qualifies when:
Does NOT qualify when:
Include author trust for every non-maintainer item:
# Quick trust check
gh api users/<login> --jq '{login, created_at, public_repos}'
Output format:
Trust: @login; acct 2023-04-03; repo 2 PRs/1 issue; activity: low/medium/high
Return structured triage output suitable for the orchestrator:
Cycle #42 — Triage Report
Repo: owner/name
Scanned: 12 issues, 5 PRs
Autonomous (3):
- GH-123: Bug — login timeout on slow networks
Type: bug | Fit: good | Risk: low
Proof: repro in description, failing test in CI
Trust: @contributor; acct 2024-01; previous merged PRs: 2
→ Dispatch to thread
- GH-124: Deps — update lodash to 4.17.21
Type: dependency | Fit: good | Risk: low
Proof: green CI
→ Dispatch to thread
Needs Review (1):
- GH-125: Feature — dark mode toggle
Type: feature | Fit: mixed | Risk: medium
→ Flag for maintainer
Defer/Close (2):
- GH-126: Stale (8mo no activity)
- GH-127: Duplicate of GH-120
current (default): triage only the current repoall: triage all repos in AGENT_LOOP_REPOS or detected from ghorg: triage all repos under the current org# Find repos with open issues/PRs
gh search repos --owner <org> --sort updated \
--json nameWithOwner,openIssuesCount,openPullRequestsCount
For each repo with open items, run the standard triage pass.
Limit to top 10 repos by activity unless AGENT_LOOP_DEEP_SCAN is set.
npx claudepluginhub duyet/codex-claude-plugins --plugin agent-loopGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.