The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, such as task definition files and source code from the repository, to generate and verify compliance checklists.
Ingestion points: SKILL.md instructs the agent to "From each task file, extract every: Spec scenario (WHEN/THEN), Done When criterion, Explicit instruction" and "read the actual code in the repository".
Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within the task files or code being analyzed.
Capability inventory: The skill utilizes file-reading capabilities to access the repository and task files.
Sanitization: There is no evidence of content sanitization or validation for the instructions extracted from external task files.

task-compliance