The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external organizational and project data provided by users in multiple formats.- Ingestion points: As described in instructions.md, the assistant gathers project data, file paths, or parameters from the user in CSV, Excel, or JSON formats.- Boundary markers: The skill instructions contain explicit constraints requiring the agent to 'Only use data provided by the user or referenced in the skill' and to 'Validate inputs before processing'.- Capability inventory: The claw.json manifest requests filesystem permissions, and the SKILL.md documentation provides examples of writing analysis reports to the local file system.- Sanitization: The instructions mandate input validation before processing, which serves as a mitigation for processing untrusted content.

data-silo-detection