The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches the SpiderFoot framework source code from a well-known repository on GitHub (github.com/smicallef/spiderfoot) during the installation phase.
[COMMAND_EXECUTION]: Uses the command line to clone the repository, install dependencies via pip, and execute the tool's main entry point (sf.py).
[COMMAND_EXECUTION]: Employs Python's subprocess.run in SKILL.md to programmatically invoke the SpiderFoot CLI with parameters derived from function inputs.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it aggregates data from over 200 external sources and presents it to the agent.
Ingestion points: Data enters the context via the parse_sf_output function in SKILL.md, which reads JSON files containing external scan results.
Boundary markers: The skill does not implement delimiters or instructions to treat the ingested OSINT data as untrusted.
Capability inventory: The skill possesses capabilities to execute shell commands (subprocess.run) and perform network requests (requests.get/post).
Sanitization: No sanitization or filtering is applied to the values retrieved from external data sources before they are processed by the agent.

spiderfoot