The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/check_health_files.py uses subprocess.run to call the gh CLI for repository health verification.
Evidence: The call subprocess.run(["gh", "api", ...]) occurs at line 125 of the Python script.
Analysis: The command is executed using a list of arguments, which prevents shell injection attacks. The interaction is limited to fetching repository metadata from a well-known service (GitHub).
[PROMPT_INJECTION]: The skill workflow involves the agent reading local repository files to customize documentation, creating an indirect prompt injection surface.
Ingestion points: SKILL.md instructs the agent to "Inspect the repo first" in step 1 of the workflow to identify languages and build commands.
Boundary markers: None are specified to protect against malicious instructions embedded in the repository files being analyzed.
Capability inventory: The skill allows writing new files to the filesystem and executing the provided verification script.
Sanitization: No sanitization of the repository content is performed before the agent processes it to generate new documentation.

community-health