Devils Advocate Idea Reviewer

Devils Advocate Idea Reviewer

Advanced operational skill for devils-advocate-idea-reviewer.

## Activation trigger

Use this skill when the user requests **devils advocate idea reviewer** and needs a high-confidence, decision-ready outcome in **idea-validation**.

## Required inputs

- Audit target (repo/module/system) and risk appetite.

• Severity model (P0-P3) and remediation SLA expectations.

• Security or quality policies that must be enforced.

• Evidence sources (code, configs, logs, prior incidents).

  Optional inputs

  • Previous outputs from this skill family.
  • Team ownership map and delivery timeline.
  • Explicit constraints for cost, risk, or compliance.

  Files to inspect

  • concept/, templates/, examples/, scripts/ in this skill folder.
  • User-referenced repository files and related modules.
  • Prior artifacts that constrain or inform this decision.

  Execution workflow

  1. Inspect the target and collect objective evidence per risk domain.

2. Classify findings by impact, exploitability/probability, and blast radius.

3. Produce remediation plan with sequencing, owners, and verification steps.

4. Flag quick wins vs structural fixes and quantify residual risk.

5. Deliver a release recommendation (blocker/warn/pass) with rationale.

   Generated artifacts

   • outputs/audit-report.md

• outputs/remediation-plan.md

• outputs/retest-checklist.md

  Output contract

  Final response must include:

  • Objective and scope boundaries.
  • Inputs and assumptions used.
  • Analysis and decision rationale.
  • Artifact paths and summary.
  • Validation result and residual risks.
  • Next actions ordered by priority.

  Validation checklist

  • Required sections are complete and non-empty.

• No placeholder content (TODO, TBD, lorem, placeholder).

• Claims are traceable to provided inputs or inspected files.

• Output includes explicit decisions, risks, and next steps.

  Safety / failure rules

  • Pause and ask for clarification if required inputs are missing or contradictory.

• Do not invent metrics, user evidence, or repository facts.

• Do not modify unrelated files or broaden scope silently.

• If high-risk uncertainty remains, return a gated recommendation instead of false precision.

  Example commands

  $devils-advocate-idea-reviewer "Run devils advocate idea reviewer on my current project context"
  python scripts/run.py --input examples/input-example.md
  python scripts/validate.py --file examples/output-example.md
  

  Advanced usage

  • Run in phased mode: discovery -> draft -> validation -> final.
  • Compare two decision branches and include tradeoff table.
  • Enforce stricter gates for production/release-critical use.