The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses git for repository analysis (history, diffs, and bisecting) and pytest for executing tests. These commands are essential for the workflow and are used in a standard, expected manner for debugging.
[PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection as it is designed to analyze external data such as stack traces and git logs.
Ingestion points: The skill reads git diffs, error messages, and test output from the codebase.
Boundary markers: No explicit instructions are provided to separate or sanitize ingested data from the agent's instructions.
Capability inventory: Filesystem access via git, code execution via pytest, and dependency management via pip.
Sanitization: There are no documented mechanisms for sanitizing or validating the untrusted content before analysis.

regression-root-cause-analyzer