The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/check-git.py uses subprocess.run to execute git merge-base commands. This is used safely by passing a list of arguments rather than a shell string, and the input command is validated to exclude shell-active characters like $, `, <, and >. The shell script scripts/block-dangerous-git.sh acts as a simple wrapper for the Python execution.\n- [INDIRECT_PROMPT_INJECTION]: The skill acts as a filter for untrusted command strings that may be generated by an agent.\n
Ingestion points: The main() function in scripts/check-git.py reads command data from sys.stdin via json.load.\n
Boundary markers: None explicitly defined in the input schema; the script is designed to serve as the security boundary itself.\n
Capability inventory: The script performs read-only Git ancestry checks using subprocess.run and provides a blocking mechanism via exit codes. The test suite scripts/test_check_git.py executes the hook script to verify the exit status contract.\n
Sanitization: The skill employs shlex for POSIX-compliant command parsing and strictly blacklists characters that allow shell expansion or redirection, ensuring it only processes literal command strings.

z-git-guardrails