github-incident-from-feedbacklisted
Install: claude install-skill bankielewicz/DevForgeAI
# github-incident-from-feedback — Skill Definition
This skill converts an array of `ai_analysis` recommendations (already parsed and selected upstream from a `devforgeai/feedback/ai-analysis/${STORY_ID}/...ai-analysis.json` file) into GitHub issues posted to `bankielewicz/DevForgeAI`. It owns drafting, preview, approval, and posting. The slash command `/create-incident-from-feedback` delegates here for any work that touches `gh issue create`.
**Sibling skills:** `github-incident-from-rca` (RCA-derived recs) and `github-incident-from-recommendations` (QA-derived recs). Same posting pipeline, different input shape — `ai_analysis` recs are leaner than QA recs (no `file`/`line`/`before_code`/`after_code`/`verification` fields; have `description`/`affected_files[]`/`implementation_notes`/`priority`/`feasible_in_claude_code` instead). Drafting therefore looks closer to the RCA path than the QA path.
**Embedded asset (the AI prompt that defines the issue body schema, quality bar, prohibited language, and DevForgeAI house rules):**
- `assets/templates/Github-incident-template.md` — read this in Phase 1; it's the authoritative drafting prompt. Byte-identical to the two sibling skills' templates (a framework-wide invariant, not source-specific).
**Why a separate skill instead of inline command logic?** GitHub-issue creation has irreversible side effects (issues leave audit trails on the public repo). The framework's anti-pattern guidance treats irreversible-side-effect logic as a L