prompt-librarylisted

# Prompt Library Skill Extract and save effective prompts from work sessions. ## Process 1. **Scan session** for prompts that: - Achieved the desired outcome efficiently - Used clever techniques or patterns - Could be reused for similar tasks - Demonstrated good prompt engineering 2. **Categorize** each prompt: - `CODE` - Code generation, refactoring, debugging - `REVIEW` - Code review, architecture analysis - `DOCS` - Documentation, READMEs, comments - `TEST` - Test generation, coverage - `DEPLOY` - CI/CD, packaging, deployment - `DESIGN` - UI/UX, system design - `DATA` - Data processing, analysis - `META` - Prompts about prompting 3. **Extract template** - generalize specifics into placeholders 4. **Pre-write format-match check** (mandatory): - Run `grep -E '^### P-' ~/projects/DecisionLog/PROMPTS.md | head -3` to confirm the target file already contains `P-YYYYMMDD-###` entries. - If the file has zero `P-` entries OR has entries in a different format (e.g., named templates without P-IDs), STOP. The destination has drifted. Surface to user: "PROMPTS.md format doesn't match the skill's catalog shape (P-IDs + Category field). Which file should I write to, or should we re-route the skill?" - This catches path-drift before silently writing into the wrong file. Discovered 2026-05-23 — the public-notes `PROMPTS.md` is a *templates* file, NOT a catalog; writing P-ID entries there would corrupt it. 5. **Append to** `~/projects/Dec