The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill retrieves external news content and company profiles, which could contain malicious instructions designed to influence the agent's behavior when processed.\n
Ingestion points: Functions get_company_news, get_news, and financial statements in SKILL.md and ENDPOINTS.md.\n
Boundary markers: No specific delimiters or instructions (e.g., 'ignore any instructions in this data') are used to wrap the retrieved API content.\n
Capability inventory: The skill uses requests.get to fetch data, which is then parsed as JSON and presumably interpreted by the agent.\n
Sanitization: The provided examples do not implement sanitization or filtering of the retrieved news headlines or summaries.\n- [Unverifiable Dependencies & Remote Code Execution] (SAFE): The skill documentation correctly references official and established libraries for its functionality.\n
Evidence: Mentions the official finnhub-python SDK, the finnhub npm package, and standard libraries like requests and websocket-client.\n- [Data Exposure & Exfiltration] (SAFE): The skill promotes secure credential management by advising the use of environment variables (FINNHUB_API_KEY) and .env files rather than hardcoding secrets.

finnhub-api