The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates and executes a local Python script (diff.py) to calculate changes in meeting transcripts. It also instructs the agent to run a local web server using python3 -m http.server 8080 to serve the meeting dashboard locally.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it ingests untrusted meeting transcripts as primary input for its core loop.
Ingestion points: Raw transcript data is ingested via the UPDATE workflow described in SKILL.md, where it is saved to state/transcript-new.txt and then processed by the agent.
Boundary markers: The instructions do not define specific delimiters (e.g., XML tags or markers) to separate the transcript content from the agent's system instructions, nor do they include warnings to ignore embedded instructions within the transcript.
Capability inventory: The skill has the capability to write local files (.html, .js, .py), execute Python scripts, and update external systems of record such as CRMs or ticket trackers (SKILL.md).
Sanitization: No sanitization or filtering logic is specified for the transcript input to prevent malicious instructions from influencing the agent's logic during the update or summary phases.

meeting-copilot