The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python scripts within Task 4 to generate financial charts. These scripts utilize well-known and trusted data science libraries (matplotlib, seaborn, pandas, numpy, plotly) to visualize financial data derived from previous tasks.
[EXTERNAL_DOWNLOADS]: The workflow involves fetching public financial data from SEC EDGAR and market data from established sources like Yahoo Finance or Bloomberg. These are standard operations for equity research and target trusted/well-known services.
[DATA_EXFILTRATION]: While the skill processes significant company and financial data, there is no evidence of unauthorized data transfer. Information is contained within the generated project files (.md, .xlsx, .docx) for user delivery.
[PROMPT_INJECTION]: The skill implements a robust 'Input Verification Protocol' and strict 'One Task at a Time' rules that mitigate the risk of accidental or adversarial instruction override. It focuses on analytical data processing rather than open-ended conversational logic.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (SEC filings, earnings transcripts, company websites). The risk is mitigated by the highly structured nature of the analytical tasks, which transform qualitative text into quantitative models and professional reports following rigid templates, reducing the surface area for the agent to 'obey' embedded instructions.
[DYNAMIC_EXECUTION]: Task 4 involves generating Python code for chart creation. This code is built from static templates provided in the reference files and populated with verified financial data, presenting no significant risk of arbitrary code execution from untrusted sources.

fsi-er-initiating-coverage