The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external websites and codebases, which are untrusted ingestion points. \n
Ingestion points: SKILL.md instructs the agent to audit subjects including 'Live URL' and 'Code' files. \n
Boundary markers: There are no instructions to use delimiters or ignore embedded directives when processing the audited content. \n
Capability inventory: The skill utilizes a browse tool for web access and references retrieval and database tools in references/refiner-protocol.md. \n
Sanitization: No validation or sanitization of ingested content is specified. \n- [EXTERNAL_DOWNLOADS]: The 'Audit Mode Protocol' in SKILL.md requires fetching content from external URLs provided by the user using a browse daemon. \n- [COMMAND_EXECUTION]: The skill specifies the use of a browse daemon tool ($B) with subcommands like goto, snapshot, and screenshot to interact with live web environments.

kai-taste