The Agent Skills Directory

[COMMAND_EXECUTION]: The skill templates instruct the agent to execute Python scripts using shell commands to perform data analysis.
Evidence: In assets/meeting-templates/customer-insight-analysis.md, the agent is directed to run python scripts/customer_interview_analyzer.py interview_transcript.txt.
Evidence: In assets/meeting-templates/product-feature-review.md, the agent is directed to run python scripts/rice_prioritizer.py features.csv --capacity 15.
Note: These scripts are not included in the provided skill files, indicating a dependency on external local code that cannot be verified.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data which could contain malicious instructions designed to influence the agent's behavior during the meeting simulation.
Ingestion points: The skill processes user-supplied files such as interview_transcript.txt and features.csv (referenced in the assets/meeting-templates/ directory).
Boundary markers: The instructions do not define clear delimiters or use specialized prompts to isolate the processed data from the agent's core simulation logic.
Capability inventory: The agent has the capability to execute shell commands, summarize potentially adversarial content, and generate documents based on these inputs.
Sanitization: There is no evidence of input validation or sanitization for the files being analyzed by the internal scripts.

multi-agent-meeting