The Agent Skills Directory

Local Script Execution: The skill includes a self-contained Python script, materialize_public_equity_sensitivities.py, used to generate structured financial tables. The agent is instructed to execute this script locally. The script is dependency-free and follows conservative programming practices, such as rigorous input validation and error handling, which mitigates risks associated with local command execution.
Indirect Prompt Injection Surface: As a tool for analyzing earnings transcripts and company filings, the skill naturally ingests untrusted external data. This creates a theoretical surface for indirect prompt injection. However, the skill processes this data through a deterministic materializer that focuses on extracting and sensitizing numeric values rather than executing instructions found within the text, which significantly reduces the potential for exploitation.
Data Management and Reporting: The skill generates various support artifacts (JSON, CSV, Markdown) and a final HTML report based on the analyzed financial data. These operations are performed using local file system paths without any network requests or exfiltration patterns detected. The use of standard templates and schemas ensures predictable output behavior.

scenario-sensitivity-generator