The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The documentation in SKILL.md and references/quickstart.md instructs the user to run pip install skill-seekers. This package does not belong to a trusted organization or repository, posing a risk of malicious code execution during installation.
[CREDENTIALS_UNSAFE] (LOW): The file references/upload_guide.md explicitly suggests storing the ANTHROPIC_API_KEY in plain text within shell profile files (~/.bashrc or ~/.zshrc), which is an insecure method for handling sensitive credentials.
[COMMAND_EXECUTION] (LOW): The skill relies on multiple scripts located in a cli/ directory (e.g., doc_scraper.py, package_skill.py) that are not included in the analyzed files, making their logic unverifiable.
[PROMPT_INJECTION] (LOW): The skill's primary function is to ingest untrusted data from URLs, GitHub, and PDFs (Category 8: Indirect Prompt Injection). There are no defined boundary markers or sanitization procedures mentioned to prevent malicious content within these sources from influencing the agent's behavior during the 'enhance' phase.

skill-seekers