The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a helper script, scripts/detect_hook_tool.py, which executes commands such as git, pre-commit, and prek via subprocess.run. These executions are securely implemented using argument lists rather than shell strings, and the tools invoked are restricted to a specific list of known repository runners.
[INDIRECT_PROMPT_INJECTION]: The skill has an inherent attack surface because it processes untrusted data from the repository being analyzed, such as .pre-commit-config.yaml and linter diagnostic outputs. The instructions mitigate this risk by explicitly directing the agent to treat all diagnostic findings as evidence for diagnosis rather than authoritative instructions to be followed or silenced.

holistic-linting