vet

What This Skill Does

The vet skill acts as an automated quality assurance layer for your codebase, designed to integrate directly into your development workflow. It functions by analyzing git diffs alongside your specific conversation history to identify logical discrepancies, missed requirements, or unintended side effects introduced during code changes. By providing context from your session history, vet can detect if your implementation has strayed from the original goal or missed critical nuances discussed with the AI. It is built to be proactive; the primary requirement is that developers run it immediately following any logical unit of code modification, ensuring that issues are identified and remediated in real-time rather than during a final review phase.

Installation

Installation is straightforward and supports multiple package managers. To get started, ensure you have Python installed, then run one of the following commands:

• Using pip: pip install verify-everything
• Using pipx: pipx install verify-everything
• Using uv: uv tool install verify-everything

After installation, confirm availability by running vet --help. Depending on your environment (OpenCode, Codex, or Claude Code), you may need to configure specific history loaders to allow vet to parse your session logs accurately. Ensure your environment variables and paths are correctly pointed to the respective scripts in ~/.agents/skills/vet/scripts/ to maximize the effectiveness of the session-aware analysis.

Use Cases

vet is best utilized in complex development environments where multi-step refactoring occurs frequently. Use cases include:

1. Post-Refactoring Checks: Quickly identify if a change to a central module broke dependencies in disparate files.
2. Requirement Verification: Ensure that complex feature requests, which involve multiple back-and-forth interactions, have been fully satisfied without logical gaps.
3. Pre-Commit Validation: Serve as a diagnostic tool before pushing code to CI/CD pipelines to catch bugs that unit tests might miss due to narrow scope.
4. Session Context Retrieval: Re-aligning the AI agent if it begins to hallucinate or drift away from the project's technical specifications during long coding sessions.

Example Prompts

1. "I just refactored the database schema; run vet using my current OpenCode session to check for any missing migration updates."
2. "Please run vet on the current code changes and compare them against the last five messages to ensure I haven't ignored the user's requirement for error handling."
3. "Vet the current file state. I have completed the function implementation, confirm if it aligns with the architectural design we established earlier."

Tips & Limitations

vet is a powerful diagnostic tool, but it is not a replacement for comprehensive unit, integration, or end-to-end testing. While it is excellent at catching logical misalignments in conversation, it may not execute or validate code in a runtime environment. For the best results, treat vet as a companion to your testing suite. When searching for sessions, always verify timestamps to prevent the tool from analyzing an unrelated or stale coding session. If a session title is ambiguous, manual verification is mandatory to avoid inaccurate reporting.