The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes external code provided via file paths, which creates a surface for processing untrusted data.
Ingestion points: SKILL.md instructs the agent to use Read and Grep on file paths provided in $ARGUMENTS.
Boundary markers: Absent. The instructions do not define delimiters to separate target code from agent instructions.
Capability inventory: Restricted to Read and Grep tools (read-only) as defined in the allowed-tools section of SKILL.md.
Sanitization: Absent. Content is read and analyzed directly without prior validation or filtering.

error-design