skills.rest
SirTheOracle
SirTheOracleOFFICIAL
Explore all skills by this author →
GitHub

adversarial-lite

Generate adversarial plans for small tasks using isolated Codex subagents.

Updated Apr 15, 2026
Search Tags:#planning#isolation#codex#synthesis#proposal#ai-workflow#adversarial-lite
One-click install
npx skills add https://github.com/SirTheOracle/forge-toolkit --skill adversarial-lite
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: adversarial-lite
Source: https://github.com/SirTheOracle/forge-toolkit/tree/main/codex-skills/adversarial-lite
Command: npx skills add https://github.com/SirTheOracle/forge-toolkit --skill adversarial-lite

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This skill enables fast, low-cost adversarial planning for small tasks by running two isolated Codex subagents to propose options and a synthesizer to generate a final plan.

Core Features & Use Cases

  • Two independent proposer passes (A and B) to explore alternative viewpoints.
  • A dedicated synthesizer pass (C) to consolidate proposals into a final plan.
  • Works for bounded bug fixes, small features, and isolated architectural decisions with minimal risk.

Quick Start

Provide a problem statement and source files, run the two isolated proposers, then run the synthesizer to generate the final plan.

Frequently Asked Questions about adversarial-lite

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a quick plan for a bounded bug fix?▼

To generate a quick plan for a bounded bug fix, provide a problem statement and source files to trigger two isolated Codex proposer passes and a synthesizer pass, producing a final_plan.md.

What is adversarial planning for small software features?▼

Adversarial planning for small features is an automated workflow where two isolated subagents propose alternative viewpoints, and a synthesizer consolidates them into a finalized plan for low-risk tasks.

Can I use isolated subagents for low-risk architectural decisions?▼

Yes, you can use isolated Codex subagents for low-risk architectural decisions by running two independent proposer passes to explore options, followed by a synthesizer pass to finalize the output artifacts.

What's the best way to automate isolated proposal generation for small tasks?▼

The best way to automate isolated proposal generation is running two independent Codex subagent passes in isolation, then using a dedicated synthesizer pass to consolidate the proposals into a final plan.

When should I avoid using fast adversarial planning?▼

You should avoid using fast adversarial planning for high-risk or complex tasks, as this workflow is specifically designed for bounded bug fixes and small features where a quick, low-cost plan is sufficient.

Does the synthesizer require output artifacts from both proposer passes?▼

Yes, the synthesizer requires output artifacts from both independent proposer passes (A and B) to consolidate the alternative viewpoints and generate the final_plan.md document.

Related Skills

Explore More Software Engineering Skills→
JuanMS20JuanMS20

one-three-one-rule

Structure technical decisions into a problem, three options, and a recommendation.

Community
Intermediate
crewrigcrewrig
20

developer

Make small code changes with verification and limited scope.

Official
Advanced
rethinkhealthrethinkhealth
6

ce:work

Automate end-to-end execution of work plans with quality checks and incremental commits.

Official
Advanced
kouisokouiso

ralphinho-rfc-pipeline

Orchestrate multi-agent DAG execution for RFC-driven feature development.

Community
Advanced