The Agent Skills Directory

[SAFE]: The skill is designed to interact with the Mowen platform via a local binary called mocli. The instructions define how to fetch and format activity events (likes, comments, follows) and look up related metadata like note titles and user nicknames. The behavior is consistent with the stated purpose of a discovery tool.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it retrieves and processes untrusted content from an external platform, including user-generated note titles, summaries, and nicknames.
Ingestion points: External data is ingested through the mocli disco activity command, specifically populating fields like notes[note_id].title, notes[note_id].brief, and users[uid].name in SKILL.md.
Boundary markers: The instructions recommend structural delimiters such as wrapping note titles in special brackets (『』) and using Markdown links for user names, which provides some separation between data and instructions.
Capability inventory: The skill uses the mocli command for read-only data retrieval.
Sanitization: There are no explicit instructions for the agent to sanitize or filter the external text for malicious prompt injection patterns.

mo-discover