Stats
Actions
Tags
From claude-commands
Enforces an LLM-driven, fail-soft implementation philosophy for pairv2: missing files degrade gracefully to fallback paths or LLM generation, and verifier LLM verdicts override non-LLM checks.
How this skill is triggered — by the user, by Claude, or both
Slash command
/claude-commands:pairv2-llm-driven-philosophyThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Pairv2 should fail only for bounded safety reasons (timeout/max cycles) or verifier
Pairv2 should fail only for bounded safety reasons (timeout/max cycles) or verifier judgment, not because a specific file path was missing.
When editing pairv2:
npx claudepluginhub jleechanorg/claude-commands --plugin claude-commandsGuides building production-grade AI features with LLM integration, RAG architecture, prompt engineering, and cost optimization.
Multi-agent adversarial verification framework that uses two independent reviewers to validate LLM output before publishing or deployment. Useful for hallucination detection and quality assurance.
Verifies high-stakes outputs via multi-agent adversarial review: dual independent agents apply same rubric, both must pass before shipping, with fix iterations to convergence.