The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches the third-party OSINT tool theHarvester from official and well-known sources, including the Python Package Index (PyPI), its official GitHub repository, and the GitHub Container Registry.
[COMMAND_EXECUTION]: Automates reconnaissance tasks using Python's subprocess.run and standard shell utilities to execute CLI tools and process data.
[PROMPT_INJECTION]: The skill processes untrusted data harvested from the public internet (search engines and APIs), creating a surface for indirect prompt injection where malicious content from external sources could potentially influence agent behavior.
Ingestion points: Results from theHarvester including emails, subdomains, and hostnames fetched from the web.
Boundary markers: Absent; harvested data is processed and printed without specific isolation delimiters.
Capability inventory: Subprocess execution (theHarvester), file writing (open), and network operations (curl).
Sanitization: Basic character replacement is used for filenames, and regex filters are applied to email suffixes.
[SAFE]: Implements secure practices for credential management by using placeholders for API keys and instructing users to store secrets in a separate configuration file (api-keys.yaml) rather than hardcoding them.

theharvester