The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill targets external code files for processing and modification.
Ingestion points: Untrusted data is ingested when the skill reads code files using Read, Glob, and Grep tools to identify comments.
Boundary markers: The skill includes a 'Scope guardrail' instructing the agent to never modify executable code or logic, but it lacks structural delimiters or specific 'ignore' instructions for the content within the comments being processed.
Capability inventory: The skill is authorized to use Write and Edit tools, granting the ability to modify files based on the content it reads.
Sanitization: There is no technical sanitization or validation performed on the comment strings before they are processed by the agent.
[NO_CODE]: The skill consists entirely of markdown instructions and does not include any executable scripts or binary files.

antislop-code