Route plans to multiple model families for parallel critique and deduplicated findings.
npx skills add https://github.com/HakAl/team_skills --skill collaborate
Please help me install this Agent Skill. Skill: collaborate Source: https://github.com/HakAl/team_skills/tree/main/collaborate Command: npx skills add https://github.com/HakAl/team_skills --skill collaborate
Cross-model review breaks cognitive monoculture by sending plans to different model families for independent critique, enabling diverse lenses and safer, more robust decisions.
Run /collaborate <plan-file> to begin a parallel cross-model review of your plan.
High-intent search queries and answers about installing and using this skill.
Cross-model review routes a plan and relevant context to multiple model families for independent critique. It applies diverse lenses to architecture planning, security assessments, and process improvements to improve accuracy and coverage.
Run the collaborate command with your plan file as input. The system fans out parallel provider calls using a shared prompt structure, then applies a deduplication workflow to present unique findings with corroboration status.
Yes, the deduplication workflow groups redundant feedback by root cause to present unique findings. It also tracks corroboration status, showing which independent model families identified the same issue.
Yes, cross-model review supports security assessments. It applies role-specific lenses to guide multiple providers in independently critiquing your security plan, surfacing diverse vulnerabilities through parallel feedback gathering.
You need a plan file containing your architecture or security proposal. The system requires a SKILL.md frontmatter with name and description, and supports optional scripts, references, and assets to guide the parallel review process.
wjyugutouExtract structured metadata from SKILL.md files and optional resource directories.
2389-researchCoordinate heterogeneous AI agents using a deterministic four-layer architecture.
markus41Integrate Claude, GPT, Gemini, and Ollama APIs with a pluggable provider abstraction.
compozySpawn opposite-model reviewers to critique code and produce verdicts.
