The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and process agent histories and charters, which may contain untrusted data or previous malicious instructions from past interactions. There is a potential risk that the agent could inadvertently promote such content into shared, reusable skills during the extraction process. * Ingestion points: The process involves reading all agent charters and interaction histories located within the squad's configuration directories. * Boundary markers: The instructions do not define specific delimiters or instructions to ignore potential commands embedded within the histories being analyzed. * Capability inventory: The skill utilizes file reading and writing capabilities to create new skill definitions at .squad/skills/ and to modify existing configuration files. * Sanitization: There are no explicit validation or sanitization steps mentioned to filter the content extracted from history logs before it is incorporated into new skill files.

reskill