The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Python to parse VTT formatted transcripts.
Evidence: The instructions in Step 2 specify using Python to extract timeline and speaker content from VTT files and save them as temporary text files for processing.
[DATA_EXFILTRATION]: The skill accesses external meeting transcripts via Microsoft 365 connectors.
Evidence: Step 1 involves searching the M365 Outlook calendar and retrieving a meetingTranscriptUrl which is then read using the read_resource tool. The data is processed locally to update files like MEMORY.md and Daily Notes.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the processing of untrusted meeting transcripts.
Ingestion points: Meeting transcripts fetched from the meetingTranscriptUrl (SKILL.md, Step 1).
Boundary markers: The skill does not explicitly define delimiters or instructions to ignore embedded commands within the transcript text.
Capability inventory: The agent is instructed to perform file system writes to Daily Notes, memorbs/, and MEMORY.md, and execute Python scripts for parsing (SKILL.md, Step 2 & 4).
Sanitization: No sanitization or validation of the transcript content is specified before it is parsed or used to generate summaries.

m365-meeting-note