From devils-advocate
Improve document based on devil's concerns, create versioned file, re-score, rename with residual. Clear 4-step workflow - improve, version, score, rename. Re-scores in place when user made changes outside Claude.
How this skill is triggered — by the user, by Claude, or both
Slash command
/devils-advocate:iterateThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
One cycle: improve, version, score, rename with residual.
One cycle: improve, version, score, rename with residual.
MANDATORY: Use TaskCreate/TaskUpdate per step (decide, apply, score, finalize, update). Mark in_progress/completed.
Prerequisites: devils_advocate.md must have scorecard with top gaps. Otherwise run /devils-advocate:evaluate first.
ASK:
"Top concerns by residual risk: [list top 5 from latest scorecard]
How to address them?
Which? (or mix: 'auto for #1-3, discuss #4')"
If user chose 1, 2, or 3 (AI makes changes):
<name>_v*.md<name>_v<NN>.md (no score suffix yet)If user chose 4: skip to Step 3. User's document IS current state.
Re-read document in full.
ASK: "Score in-session or standalone (claude -p)?"
claude -p with persona + documentPer concern:
If AI made changes (versioned copy exists):
Embed scorecard at end:
---
## Document Scorecard (Devil's Advocate)
**Persona**: [devil role and key bias]
**Score**: [total residual risk] (lower = better, max [total absolute risk])
| # | Concern | Risk | Score | Residual | How addressed |
|---|---------|------|-------|----------|---------------|
| 1 | [name] | [risk] | [0-100%] | [residual] | [specific text] |
RENAME: <name>_v<NN>.md -> <name>_v<NN>_<score>.md
report_v02.md -> report_v02_34.md_<score> MANDATORYProgression: report_v01_89.md -> report_v02_34.md -> report_v03_12.md
If user edited externally: no rename. Update devils_advocate.md only.
Always:
Report: "Iteration complete. Score: [old] -> [new]. Top gaps: [list]."
Stopping criteria:
Otherwise: "Continue with another /devils-advocate:iterate."
Score drops >20%: "Jolly good show! From [old] to [new] - devil's running out of ammunition."
npx claudepluginhub stellarshenson/claude-code-plugins --plugin devils-advocateProvides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.