adversarial-coding

A coding workflow in which separate agents plan the work, write the code, and test or criticize the result in repeated cycles. It can pause when a spending limit, iteration limit, or quality target is reached.

In plain words
What is it for?
Use it to build features or applications through repeated planning, implementation, testing, and review, with optional browser testing using Playwright.
Why use it?
It adds planning and review steps to coding tasks, so problems can be found before the work is considered finished. It also provides limits on cost and repetition.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/gettalon/talon-plugins/adversarial-coding
Any agent
npx skills add gettalon/talon-plugins --skill adversarial-coding
Clone the repo
git clone --depth 1 https://github.com/gettalon/talon-plugins

Made for: Claude Code, Codex.

Per session 45 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,402 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00045 $0.02402
Opus 5 $0.00023 $0.01201
Sonnet 5 $0.00009 $0.00480
Haiku 4.5 $0.00005 $0.00240

Measured 2d ago against content hash 48300832cbd9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

adversarial-coding scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

plugins/adversarial-coding/skills/adversarial-coding/SKILL.md · 308 lines

How it starts

The opening of the file, as written. The whole thing — 308 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Adversarial Coding — GAN-Style Multi-Agent Development

Inspired by Anthropic's research on multi-agent coding. Three agents work in adversarial loop: Planner expands requirements, Generator implements features, Evaluator tests and critiques. The system iterates until quality goals are met or checkpoints trigger user review.

Arguments

Parse $ARGUMENTS for:

  • <goal> — One-line goal (e.g., "build a retro game editor", "create a browser-based DAW")
  • --budget <dollars> — Max cost before auto-pause (default: 50)
  • --max-iterations <n> — Max Planner→Generator→Evaluator cycles (default: 20)
  • --quality-threshold <1-10> — Minimum Evaluator score to ship (default: 7)
  • --no-playwright — Disable Playwright testing (use tests + AI review only)

If arguments are missing, ask the user for the goal.

Architecture

┌─────────────────────────────────────────────────────────────────────┐
│                    MAIN ORCHESTRATOR                                │
│                                                                     │
│  ┌──────────┐      ┌──────────┐      ┌──────────┐                  │
│  │ PLANNER  │─────▶│GENERATOR │─────▶│EVALUATOR │                  │
│  │ (Plan)   │      │ (Code)   │      │ (Test)   │                  │
│  └────┬─────┘      └────┬─────┘      └────┬─────┘                  │
│       │                 │                 │                        │
│       │    feedback     │    bugs/issues  │                        │
│       └◀────────────────┴─────────────────┘                        │
│                                                                     │
│  State: .adversarial/                                               │
│  ├── spec.md           # Full specification                         │
│  ├── tasks.json        # Feature checklist with status              │
│  ├── evaluator-log.md  # Test results + AI review notes             │
│  └── checkpoint.json   # {cost, iterations, stuckCount, status}     │
└─────────────────────────────────────────────────────────────────────┘

Read the full file on GitHub · 308 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 308 lines · 45 tokens per session scan A 48300832cbd9

Subscribe to this mod's changes

adversarial-coding is a skill published in the GitHub repository gettalon/talon-plugins (6 stars, last pushed 5mo ago), licensed MIT. It adds 45 tokens to every session and 2,402 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

opencli-usage

Use at the start of any OpenCLI session — this is the top-level map of what opencli can do, how to discover adapters, what flags and output formats are universal, and which specialized skill to load next. Point here when an agent asks "what can opencli do?" or "how do I find the right command?".

jackwener/OpenCLI · 74 tokens

skill-creator

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends AIPex's capabilities with specialized knowledge, workflows, or tool integrations.

AIPexStudio/AIPex · 47 tokens

build-plugin

Complete plugin development workflow: build, test, icon, troubleshoot, and setup. Use when the user wants to build a plugin, create a plugin, troubleshoot issues, add icons, or install/configure plugins. Triggers on: build plugin, create plugin, develop plugin, new plugin, plugin icon, troubleshoot, debug, setup…

opentabs-dev/opentabs · 74 tokens

unicli-explorer

Create new Uni-CLI adapters by exploring websites and APIs. Use when adding support for a new site, desktop app, or service that unicli doesn't cover yet.

olo-dot-io/Uni-CLI · 38 tokens

website-explorer

Discover website capabilities from user behaviors. Learn APIs and automate workflows.

EndymionLee/PilotBrowseMCP · 17 tokens

context-engineering

Improve cloakbrowser-mcp agent instructions, skill routing, durable decisions, task context, or handoffs only when the user explicitly requests context optimization, migration, stale-guidance repair, or recoverable session handoff. Align guidance with current bridge code, tests, configuration, docs, and delivery…

swimmwatch/cloakbrowser-mcp · 82 tokens