Vet inspects git diffs and conversation history to surface issues introduced by a change or misunderstandings between requested and implemented work. It helps catch logic errors, missing context, or mismatch between an agent's output and the intended goal. The skill is a procedural guide to installing and running the vet CLI, interpreting results, and integrating it into developer workflows.
Run vet immediately after any logical unit of code changes — not batched — to proactively find issues. Use it before merging, after substantial refactors, or when multiple agents or contributors are collaborating and conversation history is available.
--base-commit, --model, --confidence-threshold), and guidance on interpreting results and updating model registries.Targets agentic and developer tooling environments: Claude Code, Codex, OpenCode, and other LLM-harnesses that can provide conversation history or run local CLI tools.
Vet is a proactive code vetting skill that runs after code changes to review diffs and conversation history. The four export scripts are session-history loaders for Claude Code, Codex, Gemini CLI, and OpenCode — they all require session-specific CLI args so failing without them is expected. The SKILL.md is thorough with clear installation, usage, and configuration instructions. No security concerns; scripts only read local session files and output JSON.
Well-structured skill from imbue-ai/vet. The export scripts are clean — no network calls, no credential handling, just local file parsing and JSON transformation. The SKILL.md documentation is comprehensive and covers multiple AI coding harnesses. Scripts failing without required args is correct behavior, not a bug.