The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from various potentially untrusted sources, such as external connected sources and reports from other agents, and write them into persistent storage. * Ingestion points: cloud_threads.read_dream_notes, user requests, connected sources, and Dreamer reports. * Boundary markers: The skill instructions require a strict JSON array format for storage, which serves as a structural boundary, although it does not provide natural language delimiters for the content itself. * Capability inventory: cloud_threads.write_dream_notes allows the skill to update persistent internal storage files. * Sanitization: The skill relies on a structured JSON schema for storage but lacks explicit instructions for sanitizing or escaping potentially malicious instructions embedded within the ingested data strings.

action-items