The Agent Skills Directory

[PROMPT_INJECTION]: The skill enables the ingestion of untrusted data from external sources (e.g., via scrapling::fetch) and its persistence into internal state (state::set), which constitutes an indirect prompt injection surface.
Ingestion points: Untrusted data can enter the agent's context through external worker calls such as scrapling::fetch or by retrieving stored values via state::get as described in SKILL.md.
Boundary markers: The skill does not implement or require the use of specific delimiters or 'ignore' instructions to separate untrusted data from the agent's execution context.
Capability inventory: The skill includes the fp::pipe function for orchestrating multiple worker steps and the state::set function for writing data to persistent storage, as documented in the functional description.
Sanitization: While the skill provides structural transformations (e.g., filter, map, pick), it does not include mechanisms for sanitizing or validating the contents of the data to prevent embedded malicious instructions.

fp