The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides code templates for managing user consent and data subject requests that ingest untrusted data without sanitization or boundary markers.
Ingestion points: The record_consent function in SKILL.md accepts arbitrary metadata, and the process_access_request function in SKILL.md processes user_id across multiple data sources.
Boundary markers: None present in the provided code examples to delimit user-supplied data from system instructions.
Capability inventory: The skill is configured with Bash, Write, and Edit tools, which could be targeted if an agent executes logic derived from unsanitized processed data.
Sanitization: No validation, escaping, or filtering of input parameters is demonstrated in the implementation examples.

gdpr-expert