The Agent Skills Directory

Data Ingestion Surface: The skill is designed to process external information such as company filings, transcripts, and market data. While this creates a surface for potential indirect prompt injection, the skill mitigates this by enforcing strict evidence labeling (e.g., 'Reported', 'Derived') and recommending the use of normalization tools to handle unstructured financial data.- Local Script Execution: The skill includes utility scripts (e.g., validate_tearsheet_json.py and build_tearsheet_markdown.py) to ensure data consistency. These scripts perform deterministic operations like JSON schema validation and Markdown rendering within the local environment without making network connections or accessing sensitive system files.- Context Preflight Management: The skill employs a preflight script to initialize and load user context. This mechanism is a standard approach for state management, ensuring the agent uses authorized context and source plans before beginning analytical tasks.- Source and Evidence Controls: Detailed protocols are included to verify data freshness and handle conflicting information from different financial sources. This focus on data integrity reduces the risk of the agent being misled by stale or contradictory external data.

company-tearsheet