The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts by interpolating a user-provided file path (<path-to-ga-process.yml>) into shell commands. This pattern is vulnerable to command injection if the agent fails to sanitize shell metacharacters in the input path during the execution of commands like python3 scripts/validate-gaps.py <path-to-ga-process.yml> and python3 scripts/generate-gaps-skill-package.py <path-to-ga-process.yml>.
[PROMPT_INJECTION]: The skill processes external, untrusted YAML data which presents an indirect prompt injection surface.
Ingestion points: The skill reads the user-specified ga-process.yml file and its path, which are controlled by the user or external environment.
Boundary markers: There are no explicit instructions, delimiters, or warnings to the agent to treat the contents of the ga-process.yml file as untrusted data or to ignore any instructions embedded within it.
Capability inventory: The skill possesses the ability to execute local Python scripts and perform file system writes, including writing to primary skill and command directories when explicitly authorized.
Sanitization: No evidence of input validation, schema enforcement, or content sanitization is described in the workflow before the data is passed to the generation and validation scripts.

gaps-generate