Zscole Adversarial Spec
- 554 repo stars
- Updated January 22, 2026
- zscole/adversarial-spec
Adversarial-spec is a Claude Code plugin that refines PRDs and technical specs through multi-LLM debate until models consensus.
About
Adversarial-spec is a Claude Code plugin that helps developers and PM-dev hybrids produce sharper specs before code lands. Rather than accepting the first draft from one model, it iteratively refines product requirements and technical specifications by debating between multiple LLMs until they converge on consensus. That fits the Validate phase and scope subphase as the primary shelf, because the output is a tighter boundary on what to build and how—but the artifact also feeds Build pm and docs work once engineering starts. With one focused plugin and solid GitHub traction, it is an intermediate tool: you should arrive with a rough PRD or feature outline worth arguing about. It complements brainstorming or writing-plans style skills by adding adversarial quality control, not by replacing your voice as the final decision maker.
- Single plugin that runs iterative multi-LLM debate on product and technical specifications
- Continues refinement until participating models reach consensus on the spec
- Targets PRDs, technical specifications, and multimodel adversarial review
- Claude Code-native loop so you stay in the agent instead of copy-pasting between chat tools
Zscole Adversarial Spec by the numbers
- Data as of Jul 25, 2026 (Skillselion catalog sync)
/plugin install zscole-adversarial-spec@zscole/adversarial-specAdd your badge
Show developers this plugin is listed on Skillselion. Paste this into your README.
| repo stars | ★ 554 |
|---|---|
| Last updated | January 22, 2026 |
| Repository | zscole/adversarial-spec ↗ |
What it does
Add adversarial-spec when you want PRDs and technical specs stress-tested by debating models until they agree before you commit engineering time.
Who is it for?
Best when you're writing PRDs alone and want structured disagreement between models to surface gaps in scope and technical assumptions.
Skip if: Skip if you already run formal human review lanes or anyone past build and only needs code review, not spec debate.
What you get
After registration, you get iteratively debated specifications with fewer unresolved conflicts before implementation begins.
- Consensus-aligned product or technical specification after iterative debate
- Surfaced conflicts and resolutions across model critiques
- Clearer scope boundaries to hand into implementation planning
By the numbers
- 1 plugin in the zscole/adversarial-spec bundle
- 519 GitHub stars on zscole/adversarial-spec
- Iterative multi-LLM debate continues until consensus per plugin description
Recommended Plugins
0oastro Kawai Girl Output StyleTransform Claude's responses into kawaii anime girl style\!3
0xbigboss Claude CodeComprehensive configuration system for Claude Code51
2lab Ai Oh My ClaudeClaude Code plugin for AI-powered iterative development loops.37
45ck Claude Sdlc PluginClaude Code plugin for SDLC workflow automation with Storybook planning hub1
70 10 Claude Code Plugins70-10's Claude Code plugins
Aaronbassett Agent FoundryMy personal plugin marketplace for Claude Code4How it compares
Multimodel spec debate plugin, not a codebase linter or Terraform integration bundle.
FAQ
Who is Zscole Adversarial Spec for?
It is for Claude Code users drafting PRDs or technical specifications who want adversarial, multi-model review without manual parallel chats.
When should I use Zscole Adversarial Spec?
Use it after a first spec draft and before heavy build work, when requirements feel plausible but untested against alternative interpretations.
How do I add Zscole Adversarial Spec to my agent?
Install the zscole/adversarial-spec plugin in Claude Code, enable its single plugin entry, supply your draft PRD or technical spec, and run the iterative debate-to-consensus flow in session.