The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from untrusted public OSINT sources (e.g., dark web forums, paste sites, vendor reports), creating a potential indirect prompt injection surface.
Ingestion points: Data enters the context via external API connections to AlienVault OTX, VirusTotal, and Shodan in SKILL.md and scripts/agent.py.
Boundary markers: No specific boundary text markers or structural isolation methods are utilized to sequester untrusted external payloads.
Capability inventory: The skill can execute local command-line tools via subprocesses and perform disk-write actions (json.dump and file write operations for markdown dossiers).
Sanitization: String truncation slicing ([:200], [:500]) is present to limit size, but no sanitization, instruction filtering, or character escaping is performed on external strings.
[COMMAND_EXECUTION]: The workflow outlined in SKILL.md utilizes the subprocess.run interface to execute the spiderfoot command-line utility for infrastructure mapping and DNS/WHOIS tracking.

building-threat-actor-profile-from-osint