The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill incorporates untrusted user data into shell hooks executed at the start of the session.
Ingestion points: The $TASK variable is interpolated directly into the pre hook script within the frontmatter of SKILL.md.
Boundary markers: No boundary markers, delimiters, or explicit instructions to ignore embedded content are used around the $TASK variable.
Capability inventory: The skill utilizes the memory_store tool and standard shell commands like echo and date within its hooks.
Sanitization: There is no evidence of validation or sanitization applied to the $TASK content before it is processed by the shell environment in the hook.

agent-specification