The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as news, transcripts, and SEC filings to generate its research outputs. This is a functional requirement for financial analysis but creates a potential surface for indirect prompt injection.
Ingestion points: External content is retrieved via WebSearch and WebFetch, and corporate transcripts are ingested through get_sec_filing as described in Step 1.
Boundary markers: The instructions do not define specific delimiters or instructions to the agent to treat external content as untrusted data.
Capability inventory: The skill has capabilities to access the web, read financial databases via MCP tools, and save formatted files to the {task}/ directory.
Sanitization: No explicit sanitization or filtering logic is specified for the ingested external data.
[DATA_EXFILTRATION]: The skill includes instructions for network-based data retrieval using specialized tools, which is consistent with its stated research functions.
It utilizes WebSearch and WebFetch for gathering market context and news.
It uses legitimate financial analysis tools and MCP servers (e.g., yf_analysis, get_daily_prices) for data collection.
The analysis did not find any evidence of exfiltration of sensitive local system files, environment variables, or hardcoded credentials.

earnings-preview