The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to inspect and verify local 'artifacts' (code or data files) as part of its experiment loop and final verification phase. This capability creates a surface for indirect prompt injection where malicious instructions embedded in the processed files could be interpreted as commands by the agent.
Ingestion points: Artifact inspection described in Section C ('Measure against the predicate on the real artifact') and Section E ('Check the entire result against the original predicate') of SKILL.md.
Boundary markers: The instructions do not define specific delimiters or warnings to ignore instructions embedded within the analyzed artifacts.
Capability inventory: The skill has access to local tools for file modification, measurement, and potentially repository management (commits).
Sanitization: There is no evidence of sanitization or content validation for the data ingested from local artifacts.
[EXTERNAL_DOWNLOADS]: The THIRD_PARTY_NOTICES.md file contains references to source code and plugins hosted on the Cursor editor's official GitHub repository. This is a well-known service used for development tool distribution.

auditable-run