The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted user files and code, creating a potential surface for indirect prompt injection.
Ingestion points: The skill instructions in SKILL.md use the Grep and Read tools to ingest content from various file types across the environment.
Boundary markers: No explicit delimiters or instructions are used to separate the ingested data from the agent's internal logic, making it possible for content in scanned files to be interpreted as instructions.
Capability inventory: The SKILL.md file enables the use of Read, Grep, and Glob tools and utilizes Python libraries like numpy and pandas for data processing.
Sanitization: There is no evidence of sanitization or filtering applied to external file content before it is analyzed or reported.

pattern-detection