The Agent Skills Directory

[SAFE]: The skill workflow is restricted to read-only repository analysis and metadata evaluation. It leverages git commands to identify code changes and coordinates with internal sub-skills to perform risk and test quality assessments. No indicators of data exfiltration or unauthorized access were found.
[SAFE]: Explicit prohibitions are included to prevent the agent from performing destructive actions or modifications to the repository, such as committing or pushing code. This maintains the integrity of the development workflow.
[SAFE]: The skill analyzes local files and documentation to perform its audit functions. While processing external content presents a theoretical surface for indirect prompt injection, the implementation focuses on legitimate metadata extraction and compliance checking within a trusted local environment.

pre-pr-audit