The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code and content from external repositories, creating a vulnerability surface where malicious payloads in the audited data could attempt to influence the agent's behavior.
Ingestion points: The skill uses Read, Glob, and Grep tools to ingest content from the directory path provided in the <repository-path> argument.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' directives to separate the content being audited from the agent's internal instruction context.
Capability inventory: The skill is configured with Write and Edit capabilities, allowing it to modify files in the local environment based on its analysis.
Sanitization: There are no specific mechanisms described for the skill to sanitize or filter the ingested code before analysis, although it does instruct the agent to check for such sanitization in the audited code itself.

firecrawl-known-pitfalls