The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Python subprocess module to interact with the local environment for configuration and formatting tasks.
scripts/initialize.py executes uv tool list to verify if the rumdl markdown formatting utility is available via the uv tool manager.
scripts/_try_rumdl.py invokes the formatting command rumdl fmt on generated review files. The command execution is implemented using shlex.split and list-based arguments without shell=True, which is a secure practice for preventing shell injection.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it is designed to ingest and analyze external code as part of its primary review function.
Ingestion points: The agent reads target code files as specified in the review workflow described in SKILL.md.
Boundary markers: The skill's scripts and templates do not currently enforce specific boundary markers or 'ignore' instructions for the data being processed.
Capability inventory: The skill has the capability to write files (reports) and execute a local formatter via subprocess.
Sanitization: The validation scripts (validate_detail.py, validate_summary.py) focus on structure and placeholders rather than sanitizing the content of the review targets for adversarial instructions.

iceberg-code-review