The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python's subprocess module to execute git commands (e.g., git rev-parse, git symbolic-ref, git remote) across several scripts (install_hooks.py, file_size_hint.py, git_sync_policy.py). These calls are used for environment discovery, such as resolving the repository root and identifying the default branch.
[PERSISTENCE]: The installation script (scripts/install_hooks.py) modifies the local .git/hooks/post-commit file and registers tool-use hooks in .codex/hooks.json and .cursor/hooks.json. These hooks ensure specific Python logic (file size monitoring and Git sync policy enforcement) runs automatically during Git events and agent tool invocations.
[DYNAMIC_EXECUTION]: The skill generates shell launcher scripts and configuration files (.agents/hooks/cleanup, .mergify.yml) at runtime. It also uses ast.literal_eval to parse tool inputs in hooks/file_size_hint.py. These behaviors are used to bridge the agent environment with the local shell and are consistent with the skill's configuration purpose.
[INDIRECT_PROMPT_INJECTION]: The skill explores repository content, including workflow files in .github/workflows/, to suggest configuration settings. This ingestion of potentially untrusted data from the repository serves as an attack surface for indirect prompt injection.
Ingestion points: Reads .github/workflows/ and repository metadata via git remote and gh pr list commands to determine merge and hook requirements.
Boundary markers: No explicit instruction delimiters are used in the prompt when interpolating repository data, though the skill requires human confirmation before writing files.
Capability inventory: Includes file system writing (write function), permission modification (chmod), and arbitrary shell command execution via subprocess.run on Git binaries.
Sanitization: Employs shlex.quote in scripts/install_hooks.py to sanitize paths when generating the Git hook launcher shell script.

setup-harness