The Agent Skills Directory

[PERSISTENCE]: The skill instructions in SKILL.md explicitly suggest that the user or agent configure a cron job (e.g., 0 9 1 * * python3 ...) to run the update check script automatically every month. This is a persistence mechanism that allows the skill's code to maintain a presence and execute on a schedule without manual intervention.
[COMMAND_EXECUTION]: The included script scripts/check_updates.py utilizes subprocess.run to invoke system-level git commands (such as ls-remote and rev-parse) to interact with remote and local repositories. While these are necessary for the skill's primary function of update checking, they involve the execution of shell-level operations.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests and processes metadata from external files located within the skills directory.
Ingestion points: The scripts/check_updates.py script reads and parses .huashu-skill-meta.json files from various subdirectories to identify repository URLs and commit hashes.
Boundary markers: No specific delimiters or safety instructions are used to separate the metadata content from the agent's processing logic.
Capability inventory: The skill can execute Git subprocesses, read and write local files (e.g., .last-update-check), and interpolate data from metadata files into the agent's output context.
Sanitization: The script relies on standard JSON parsing but does not implement robust validation or sanitization for the repo or subdir fields before using them in command arguments or displaying them to the user.

huashu-skill-updater