| name | triage |
| description | This skill should be used when the user asks to "triage", "what needs fixing", "what's broken", "prioritize issues", "run triage", or wants a P0/P1/P2 issue breakdown across one or more repos. |
Triage Workflow
Scan HANDOFF files and GitHub issues, categorize by priority, and spawn parallel subagents
for fixes.
Steps
1. Read current state
ls .ctx/HANDOFF*.yaml 2>/dev/null || ls HANDOFF*.yaml 2>/dev/null
Read each HANDOFF.*.yaml found. Focus on items with status: open or status: blocked.
2. Check GitHub issues
gh issue list --repo <repo> --state open --limit 50 --json number,title,labels,assignees
Use the repo dirname (not 'workspace') as <repo>. Do NOT use Linear — GitHub is the issue
tracker for this workspace.
3. Categorize P0/P1/P2
| Priority | Criteria |
|---|
| P0 | broken, fails, blocked, urgent, security, CI red |
| P1 | specific file + known fix, test failing with known cause |
| P2 | enhancements, cleanup, anything safe to defer |
Merge HANDOFF items and GitHub issues into a single prioritized list. Present summary:
## Triage — <repo>
P0 (act now):
- [id] <title> — <why urgent>
P1 (autonomous fix):
- [id] <title> — <file> → <known fix>
P2 (delegate / defer):
- [id] <title>
4. On user selection, execute
- P0: Validate state, ask go/no-go before acting.
- P1: Fix autonomously. Stop if scope expands beyond the described fix or >3 files change.
- P2: Spawn one subagent per item (cap 5 concurrent). Each subagent must: receive explicit
--allowedTools, verify clean git status, commit its own changes.
5. Validate with nextest
cargo nextest run
cargo nextest run <filter>
Use cargo nextest, not cargo test, for test filtering and parallel execution.
6. Commit with handoff updates
After fixes:
git add .ctx/HANDOFF.*.yaml
git commit -m "fix: <summary> — closes #<issue>"
Update HANDOFF items to status: done and prepend a log entry. Do not edit HANDOFF manually
for status if doob sync is active — use handoff-db upsert instead.
