Claude Skill MartBrowse skillsQuick linesLearn by videoTerminal guideWhat is a Skill?
 
← Back to list

Adversarial Spec

Refines a PRD or technical spec through multi-round debate between Claude and external LLMs (GPT, Gemini, Grok, and more) until every participant agrees.

Docs & OfficeAdvanced★ 553⑂ 48AI score 8/10Last updated: Jan 22, 2026

What it does

Turns a rough idea or existing draft into an implementation-ready specification by pitting several models against each other.

  • Two document tracks: PRD (problem, personas, user stories, KPIs, in/out scope) or Technical Spec (architecture, API contracts, data models, security, observability, deployment).
  • Optional interview mode: a deep requirements session across 8 topic areas (problem, users, functional requirements, constraints, UX, tradeoffs, risks, success criteria) before any debate starts.
  • Parallel multi-provider critique: via litellm (OpenAI, Anthropic, Google, xAI, Mistral, Groq, DeepSeek, OpenRouter, Zhipu), the codex/gemini CLIs, or entirely through AWS Bedrock for compliance-bound teams.
  • Anti-laziness check: if a model returns [AGREE] in rounds 1–2, the --press flag forces it to prove it read the whole document.
  • Power features: --focus areas, professional --personas, --context injection of existing APIs/schemas, named sessions with resume, per-round checkpoints, --preserve-intent to stop convergence from sanding off deliberate choices, token/cost reporting, and Telegram human-in-the-loop.
  • On consensus the final doc is written to spec-output.md, with an optional PRD → Tech Spec continuation.

Who it's for

  • Solo founders and small-team PMs with no one to pressure-test a spec.
  • Tech leads who want design gaps and edge cases surfaced before a sprint starts.
  • Enterprise teams that must route all model calls through AWS Bedrock.

Examples

  1. "Draft a PRD for payment reconciliation, start with interview mode" → deep interview, generated draft, then 3–4 rounds where gpt-4o and Gemini attack vague metrics and missing scope boundaries.
  2. Feed an existing ./docs/auth-spec.md with --focus security --persona security-engineer to hunt for auth, authorization, and input-validation holes.
  3. Run cycle 1 with cheap fast models for completeness, then cycle 2 with o1 / claude-opus for a final hard review.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/zscole/adversarial-spec/HEAD/skills/adversarial-spec/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the skills/adversarial-spec folder from the GitHub repo zscole/adversarial-spec into my ~/.claude/skills/adversarial-spec/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/zscole/adversarial-spec.git /tmp/adversarial-spec && mkdir -p ~/.claude/skills && cp -r /tmp/adversarial-spec/skills/adversarial-spec ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Prerequisites: install Python 3.10+ and run pip install litellm. Do NOT install the unrelated llm package.
  2. Set up auth: export at least one provider key, e.g. export OPENAI_API_KEY=sk-... or export GEMINI_API_KEY=.... For subscription access instead: npm install -g @openai/codex && codex login, or npm install -g @google/gemini-cli && gemini auth.
  3. Install the skill: 
    git clone https://github.com/zscole/adversarial-spec.git /tmp/adversarial-spec
mkdir -p ~/.claude/skills
cp -r /tmp/adversarial-spec/skills/adversarial-spec ~/.claude/skills/
  4. Verify providers: run python3 "$(find ~/.claude -name debate.py -path '*adversarial-spec*' | head -1)" providers to see which keys are detected.
  5. Fix auth conflicts: if you get the "both a token and an API key are set" error, either unset ANTHROPIC_API_KEY or run claude /logout.
  6. Try it: restart Claude Code and ask something like "refine my PRD using adversarial debate" — the skill will ask for document type and opponent models.
  7. (Optional) Telegram: create a bot with @BotFather, export TELEGRAM_BOT_TOKEN and TELEGRAM_CHAT_ID, then add the --telegram flag for round-by-round notifications.
View source on GitHub ↗License: MIT