zscole/adversarial-spec
A Claude Code plugin that iteratively refines product specifications by debating between multiple LLMs until all models reach consensus.
The plugin uses `litellm` to orchestrate parallel critique cycles across 10+ LLM providers (OpenAI, Anthropic, Google, xAI, Mistral, etc.), enabling cost-effective multi-model comparison through unified API abstraction, OpenRouter gateway, or AWS Bedrock for enterprise compliance. Claude participates as an independent critic rather than orchestrator alone, contributing substantive revisions and synthesizing feedback until convergence before the user review phase. Architecture supports flexible model routing including Codex/Gemini CLIs for subscription-based access, with configurable reasoning effort levels for complex analysis.
509 stars.
Stars
509
Forks
43
Language
Python
License
MIT
Category
Last pushed
Jan 22, 2026
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/llm-tools/zscole/adversarial-spec"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
jahwag/ClaudeSync
ClaudeSync is a Python tool that automates the synchronization of local files with Claude.ai Projects
ducan-ne/opencoder
The Claude Code alternative
ggozad/haiku.skills
Skill-powered AI agents implementing the Agent Skills specification with pydantic-ai
SynaLinks/synalinks-skills
Claude skills for Synalinks OSS
selftune-dev/selftune
Skill observability & self-improving toolkit for agent skills.