The Agent Skills Directory

[SAFE]: The skill incorporates robust safety protocols, including a refusal to interact with databases or services not explicitly verified as disposable (e.g., localhost) and the use of 'git stash' to create recovery points before any code mutations are performed.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to discover and run testing frameworks (e.g., npm test, pytest, make). Commands are derived from standard project configuration files found in the target repository, following existing development conventions.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes repository contents like READMEs and docstrings to determine expected code behavior. While processing external data is a potential attack surface, the skill mitigates risk by requiring external validation for expectations and maintaining strict boundaries on which files it can modify.

testkit