Skip to content

ayagmar/adversarial-plan-approval

v0.1.1MIT

Independently falsify or approve implementation plans before execution with ACCEPT, MODIFY, or REJECT verdicts per task or decision gate.

GitHub

adversarial-plan-approval

Review, challenge, or approve an implementation plan before work starts. Privacy: instruction-only skill; ships no scripts, reads no session stores, and makes no network requests of its own.

Requirements

  • None beyond a coding agent that supports Agent Skills

Install

Pi

# pinned (versions are listed in the changelog linked below)
pi install npm:@ayagmar/adversarial-plan-approval@<version>

# unpinned
pi install npm:@ayagmar/adversarial-plan-approval

# remove
pi remove npm:@ayagmar/adversarial-plan-approval

skills CLI

npx skills add ayagmar/agents-skills --skill adversarial-plan-approval --agent pi

Other agent ids: claude-code, codex, cursor.

Claude Code

claude plugin marketplace add ayagmar/agents-skills
claude plugin install adversarial-plan-approval@ayagmar-skills

Codex

codex plugin marketplace add ayagmar/agents-skills
codex plugin add adversarial-plan-approval@ayagmar-skills

Usage

Prompt example: "Review plans/003-retry-policy.md before we start and give me ACCEPT, MODIFY, or REJECT for each task." The agent loads the skill when your request matches its description. See SKILL.md for the full workflow.

Links