The Agent Skills Directory

[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it processes external, potentially untrusted markdown transcripts from Google Meet.
Ingestion points: Transcripts are ingested via the Read tool from the Meeting notes/ directory (specifically files matching *회의록.md).
Boundary markers: No explicit delimiters or boundary markers are defined to isolate transcript content from the agent's instructions.
Capability inventory: The skill is limited to file system discovery (Glob), pattern matching (Grep), file reading/writing (Read, Write), and user interaction (AskUserQuestion). It does not possess network access or arbitrary command execution capabilities.
Sanitization: The skill does not perform sanitization or escaping of the ingested text before summarization.
Mitigation: The skill requires a 'User Review' phase using AskUserQuestion before any Write operations, ensuring a human-in-the-loop validates the output before it is persisted to disk.

meeting-notes