The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for ingesting and processing large amounts of untrusted data from external sources, which could contain adversarial instructions. * Ingestion points: Untrusted external data sources such as WHOIS, DNS records, public repositories, and breach databases listed in the 6-stage recon pipeline (SKILL.md, Section 7). * Boundary markers: The methodology utilizes a structured schema for findings (Section 3) and limits raw HTTP captures to 2 KiB (Section 4). * Capability inventory: The skill describes the use of external tools for active probing, such as Nuclei, nmap, and various API discovery techniques (SKILL.md, Section 7 and 11). * Sanitization: Instructions include redacting passwords from reports and encrypting stealer logs at rest (Section 12).
[SAFE]: The skill incorporates strong defensive practices, such as mandatory authorization checks (Section 1), detectability tagging (Section 6.2), and strict prohibitions on pasting real PII or credentials into cloud-hosted LLMs (Section 5).

osint-methodology