The Agent Skills Directory

[Category 4: External Downloads] (SAFE): The skill contains developer instructions for setting up the environment using standard package managers. This includes npm install for the frontend and npx for executing MCP servers. These are appropriate for the skill's stated purpose of assisting with GPT Researcher development.
[Category 8: Indirect Prompt Injection] (SAFE): The research agent described (GPT Researcher) is designed to ingest untrusted web and local data, which represents a known attack surface for indirect prompt injection. As a documentation skill, it outlines the capability without introducing specific vulnerabilities in the instruction set itself.
Ingestion points: Web search results and local file paths (specified in SKILL.md usage examples).
Boundary markers: None explicitly shown in the code snippets.
Capability inventory: Subprocess execution through MCP (npx), file reading for local research, and network access for web crawling.
Sanitization: Not detailed in the provided snippets, though common error handling patterns are mentioned.
[Category 2: Data Exposure] (SAFE): The skill demonstrates safe practices for credential management by using os.getenv("GITHUB_TOKEN") rather than hardcoding sensitive keys.

gpt-researcher