The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it is designed to ingest and process untrusted external data.
Ingestion points: The instructions.md file directs the agent to gather and process input data from users in CSV, Excel, and JSON formats.
Boundary markers: There are no explicit instructions to use delimiters or ignore embedded instructions within the processed data files.
Capability inventory: The claw.json manifest requests filesystem permissions, which are used to read project data and export reports as described in the instructions.
Sanitization: The skill instructions include a general constraint to "Validate inputs before processing," which provides a basic check but does not specify robust sanitization or escaping of external content.

capacity-planning