The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external application bundles and binaries, which presents an inherent surface for indirect prompt injection via file metadata. However, this is necessary for its diagnostic purpose and is considered safe in this context.
Ingestion points: The app argument in SKILL.md allows the agent to read external files.
Boundary markers: None explicitly defined.
Capability inventory: Uses read-only system tools (codesign, spctl, plutil) for inspection.
Sanitization: None described.
[COMMAND_EXECUTION]: The skill instructs the agent to use standard macOS utilities to inspect signing info and entitlements.
Evidence: Uses codesign -d, spctl, and plutil in SKILL.md.

fix-codesign-error