The Agent Skills Directory

[SAFE]: The skill consists entirely of instructional methodology for security reconnaissance and does not include any scripts, executables, or code that runs on the agent's host.
[EXTERNAL_DOWNLOADS]: The skill documentation suggests the use of well-known third-party OSINT services and tools (e.g., Shodan, Censys, SecurityTrails, HudsonRock). It also mentions the manual installation of command-line tools like 'cero' via Go for the human operator, which is standard for the security research context.
[INDIRECT_PROMPT_INJECTION]: While the skill involves the ingestion of external data from websites and mobile apps, it provides explicit guidelines to verify findings across multiple sources and strict instructions to redact sensitive information, such as PII or credentials, to prevent them from being sent to cloud-based large language models.
[DATA_EXFILTRATION]: The methodology emphasizes data hygiene and safe handling of sensitive findings, explicitly forbidding the exfiltration of credentials or session tokens to external LLM services and recommending the use of local models for sensitive analysis.

osint-methodology