The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands to perform repository discovery, file operations, and script verification.
Evidence: git rev-parse --show-toplevel, rsync -a --ignore-existing ..., cp -n ..., and python3 .github/actions/gate/run.py.
[DYNAMIC_EXECUTION]: The skill populates the repository with a Python runner script from its internal templates and then executes it during the setup flow.
Evidence: Step 2 copies run.py to .github/actions/gate/, and Step 5/6 executes it using python3.
[INDIRECT_PROMPT_INJECTION]: The skill has a potential attack surface for indirect prompt injection as it reads existing workflow files and incorporates user-provided configuration into the repository structure.
Ingestion points: Reads existing workflow files in .github/workflows/ and accepts repository 'owner' metadata from the user.
Boundary markers: None present.
Capability inventory: File system writes, shell command execution (rsync, cp, git), and Python script execution.
Sanitization: No explicit sanitization of ingested file content or user input before usage in configuration files.
[TIME_DELAYED_CONDITIONAL]: The skill uses an environment-specific trigger ('Overlay' mechanism) that changes its behavior based on the presence of local environment variables.
Evidence: The skill checks for the existence of $POWERSUIT_OVERLAY to determine if it should follow alternative instructions or load custom templates from a hidden path.

gate-init