The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to audit external skill directories provided via the --root argument. This represents an ingestion point for potentially untrusted data from third-party skill files.
Ingestion points: Files within the directory structure specified by the user or agent in the --root parameter (e.g., SKILL.md, agents/*.yaml).
Boundary markers: The skill instructions do not specify explicit delimiters or "ignore embedded instructions" warnings for the content being audited.
Capability inventory: The skill has the capability to execute shell commands via the ai-skillet tool.
Sanitization: The skill acts as a validation engine (ai-skillet doctor), which inherently involves checking and potentially rejecting malformed data, though the level of sanitization for shell injection in arguments depends on the agent platform.
[COMMAND_EXECUTION]: The skill invokes the ai-skillet CLI tool, passing user-provided arguments to perform audits and repairs. This is standard behavior for a developer utility and does not involve unauthorized command execution.

skill-doctor