The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied data, such as lab reports and portal exports, which serves as a potential vector for indirect prompt injection attacks.- Ingestion points: Processes user-pasted text (lab reports, portal exports, app backups) and shared memory files (profile.md, memory.md) from the local environment.- Boundary markers: The skill uses specific section headers (## Boxes, ## Due) and YAML configuration files to structure data, but does not implement explicit delimiters or 'ignore' instructions for interpolated user content.- Capability inventory: Performs local file system read and write operations within specifically scoped directories (~/Clawic/data/).- Sanitization: The instructions include a dedicated rule to identify and redact sensitive authentication data (tokens, logins, member IDs) from user-provided inputs before writing them to disk.

Nutrition