← ClaudeAtlas

tackle-issuelisted

Project-scoped extension of the user-level tackle-issue skill, active only inside this monorepo. Same triage workflow (read-only investigation, falsification, verdict, SHIP/CAVEAT/DEFER/SKIP recommendation), plus a picker mode when no specific issue is named, plus an optional branch + TaskCreate handoff on positive verdicts. Three input modes: • GitHub issue number — "tackle issue 99", "work issue 17", "start on gh #42". Fetches via gh, runs the full triage workflow. On SHIP / SHIP WITH CAVEAT, offers to check out a branch and seed implementation tasks. • Pasted task content — "investigate this proposal", "pressure-test this", "scope this before I touch it". Runs the triage workflow on the pasted text. Read-only output only — no branch, no TaskCreate seeding. • No input / picker phrasing — "tackle the next issue", "pick an issue", "what should I work on", "next issue". Lists ready-labelled open issues in gtapps/claude-code-hermit, filters those with linked open PRs, applies the dedup log, picks one, then runs
gtapps/claude-code-hermit · ★ 59 · AI & Automation · score 81
Install: claude install-skill gtapps/claude-code-hermit
# Tackle Issue Project-level extension of the user-level `tackle-issue` skill. Picks the next ready GitHub issue (or accepts a specific number or pasted task text), runs the full triage workflow against the current code, and — on a positive verdict — checks out a branch and seeds implementation tasks. Implementation, commit, and PR happen outside this skill. Three modes: **picker** (no arg), **issue number** (e.g. `tackle issue 99`), **pasted text** (investigation-only). ## Mindset A GitHub issue or pasted proposal is a **hypothesis**, not a spec. Issues filed by AI agents (hermit-scribe is the common case here) and pasted proposals often have correct intent but imperfect framing — the file paths may be stale, the proposed fix may be the first idea rather than the best one, the scope may be larger than needed, or the situation may have already changed since the issue was filed. Worse, the bug may not exist at all. The default posture is skeptical investigation. Before drafting any plan, try to falsify the premise. If the premise survives the falsification attempt, then evaluate whether implementing it is actually worth the cost — a verified premise is not the same as a worthwhile change. Two failure modes to avoid: - **Rubber-stamping** — accepting the issue's framing and drafting a plan that implements exactly what was written, even when a simpler fix exists or the premise is partially wrong. - **Reflex-pushback** — manufacturing objections to look thorough. If the pr