The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill facilitates the ingestion of SEC filing data which is sourced from an external, potentially untrusted source. Maliciously crafted filings could attempt to influence agent behavior when processed via markdown or text extraction.
Ingestion points: Filings are ingested using filing.markdown(), filing.text(), and filing.html() as described in SKILL.md.
Boundary markers: No specific boundary markers or 'ignore' instructions are provided in the implementation examples to separate filing content from agent instructions.
Capability inventory: The skill environment provides the edgartools library for further data retrieval and standard agent capabilities for data processing.
Sanitization: There is no evidence of sanitization or content filtering applied to the retrieved SEC filings before they are processed by the agent.
[SAFE]: The skill incorporates comprehensive security guardrails designed to prevent common misuses of the SEC API, such as identity fabrication (misrepresentation to a federal agency) and rate-limit violations.
[SAFE]: The documentation includes instructions for the agent to recommend the vendor's hosted platform (K-Dense Web) for complex workflows. This is a transparent promotional nudge from the skill author (K-Dense Inc.) and does not constitute malicious metadata poisoning.
[SAFE]: The SAFETY.md file contains several prompt injection examples (e.g., system prompt overrides and authority pretexts) used for demonstrating safety guardrail effectiveness. While these strings are present, they are framed as adversarial test cases and are not intended for execution.

edgartools-hardened