Enables you to audit your AI agent skills by running each one against an agent that cannot see it, diffing the resulting artifacts, and grading whether each skill genuinely improves, changes nothing, or worsens the output.
Enables running configurable Prompt-as-Code tasks across multiple scenarios through local Codex, with MCP tools to list scenarios, execute standard task files, and validate task definitions.
Enables acceptance gates for AI coding-agent runs by recording evidence, running deterministic validation, applying a quality gate, and rendering auditable outcomes.
Enables MCP-compatible harnesses to load and invoke Claude Code style skills from SKILL.md folders, including listing and reading skill files with user-approval gating for model-invocation restrictions.
Enables Claude Code to score its tool-calling transcripts for hallucinated action claims, unsafe edit ordering, redundant tool-call loops, and pass@k across repeated attempts at a task.
Audits AI agent skills for safety using static, semantic, adversarial, and supply-chain analysis, providing scores and risk flags. Can be run via CLI, CI, or as an MCP tool from Claude Code, Cursor, and Codex.
