Triage

Instructions

Step 1 — Resolve GitHub repo

Determine the target repo for the GitHub issues check:

1. Read agent-setup.yml from the project root. If a top-level github_repo: key is present, use its value as the repo (e.g. owner/repo).
2. If no github_repo: key is found (or the file does not exist), run:

   git remote get-url origin 2>/dev/null
   

   Parse the output to extract owner/repo — handle both HTTPS (https://github.com/owner/repo) and SSH (git@github.com:owner/repo.git) formats.
3. If neither source yields a repo, set repo to (unresolved) and continue — Step 2 will skip gracefully.

Step 2 — Fetch open GitHub issues (non-blocking)

If a repo was resolved and gh is available, run:

gh issue list --repo <resolved-repo> --state open 2>/dev/null

Collect all open issues (number, title, labels).

• If no repo was resolved, print GitHub issues check skipped — no repo configured. and continue.
• If gh is not installed, print GitHub issues check skipped — gh not available. and continue.
• If the command fails for any reason, print GitHub issues check failed — continuing. and continue.
• Never block or halt on this step.

Step 3 — Read backlog

Check whether docs/backlog.md exists. If it does, read the full file and note every item with its section heading. If the file does not exist, treat the backlog as empty and continue.

Step 4 — Cross-reference, deduplicate, group by theme

Compare the GitHub issues list from Step 2 against the backlog items from Step 3:

1. Deduplicate — if a backlog item and a GitHub issue clearly describe the same work (same domain + intent, even if worded differently), treat them as one candidate. Keep the GitHub issue number as the canonical reference; note the backlog item alongside it.
2. Group by theme — assign each candidate to a theme bucket based on its domain or subject area. Common themes: observability, skill-sync, enforcement, sprint-workflow, agent-authoring, documentation, infrastructure. Add new theme buckets as needed — do not force items into mismatched buckets.
3. Prioritize within each theme — lead with stop-and-fix items (active drift that gets worse every sprint), then missing-improvement items (value-add that does not degrade over time).

Step 5 — Emit triage report to chat

Output the triage report directly to chat. Do not write it to disk.

Format:

## Triage Report — <today's date>

### <Theme>
- [<source>] #<N> or B<N> — <title> — <one-sentence why it matters>
- ...

### <Theme>
- ...

---
Sources: GitHub issues (<resolved-repo> or skipped) + docs/backlog.md (<item count> items or absent)

Tag each item with its source: [issue] for GitHub issues, [backlog] for backlog-only items, [both] for deduplicated cross-references.

If both sources are empty or unavailable, emit: Triage complete — no open issues and no backlog items found.

Step 6 — Recommend sprint candidates (read-only)

After the report, emit a short recommendation block:

## Sprint Candidate Recommendations

Recommended for next sprint (stop-and-fix first):
1. ...
2. ...
3. ...

Queued (no active degradation):
- ...

This recommendation is advisory only. No changes are made to docs/backlog.md or any other file. Backlog promotion happens only when Tim explicitly confirms.

Hard Constraints

• NEVER edit docs/backlog.md — this skill is output-only. Any change to docs/backlog.md requires explicit Tim confirmation and must happen outside this skill.
• Input sources: GitHub issues + docs/backlog.md only. No Teams, Slack, or external connectors.
• Report is emitted to chat, not written to disk.
• This skill never blocks or halts due to unavailable external sources — all external checks degrade gracefully.