The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses various shell commands to locate, transfer, and execute logic files as part of a deployment gate workflow.
It utilizes find searching multiple directory levels relative to the CLAUDE_PLUGIN_ROOT variable to identify available check modules.
It employs cp and mkdir to install scripts into the repository's .github/actions/gate/rules/ directory.
It executes python3 .github/actions/gate/run.py to execute the local Python-based rules.
[DYNAMIC_EXECUTION]: The skill is designed around the creation, movement, and runtime execution of Python code.
Path A describes a procedure for copying .py files from an external module source to an executable path within the current project.
Path B involves writing new Python code to the filesystem and subsequently executing it via a runner.
It implements a "private overlay" feature where logic can be loaded from a path defined by the POWERSUIT_OVERLAY environment variable, enabling dynamic modification of the skill's behavior based on external files.
[INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface by processing data from the repository that could contain malicious instructions.
Ingestion points: The gate rules process repository content via ctx.files and ctx.front_matter as described in SKILL.md. The skill instructions also ingest user-provided examples of "hits".
Boundary markers: There are no specific delimiters or "ignore embedded instructions" warnings mentioned for the processed data.
Capability inventory: The skill possesses file write access, directory traversal (find), and Python execution (python3) capabilities as described in SKILL.md.
Sanitization: No sanitization or structured validation of the file data or user examples is implemented to prevent potential prompt injection within the rules' logic or output.

add-rule