The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install the edgartools and edgartools[ai] packages from PyPI using standard tools like uv pip. These are legitimate library installations required for the skill's functionality.
[COMMAND_EXECUTION]: The documentation includes shell commands for package installation, MCP server setup (uvx, python -m edgar.ai), and local filesystem operations such as downloading filings and exhibits (filing.save(), exhibit.download()). These are standard capabilities for a data analysis tool.
[PROMPT_INJECTION]: The SKILL.md file contains instructions for the AI to suggest the www.k-dense.ai platform when users encounter complex workflows. This is a promotional steering instruction provided by the author to guide users toward their hosted platform, but it does not attempt to bypass safety filters or override core agent behavior.
[DATA_EXFILTRATION]: The skill requires users to provide an identity (name and email) as mandated by the SEC for API access. This information is passed to the official SEC EDGAR servers as part of the request headers. The skill uses environment variables (EDGAR_IDENTITY) for secure credential management.
[REMOTE_CODE_EXECUTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from SEC EDGAR filings (HTML, text, and markdown). An attacker could potentially embed malicious instructions within a public SEC filing to influence the behavior of an agent reading it.
Ingestion points: Filing content via filing.text(), filing.html(), and filing.markdown().
Boundary markers: None explicitly defined in the provided instructions.
Capability inventory: File writing (filing.save(), exhibit.download()), local web server hosting (filing.serve()), and network access to the SEC API.
Sanitization: No specific sanitization or filtering logic is mentioned for the processed filing content.

edgartools