recon-js-analysis

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill describes a methodology for ingesting and analyzing external JavaScript files. This presents a theoretical surface where malicious instructions embedded in target code could influence agent behavior, though the skill itself is purely informational.
  • Ingestion points: Target JavaScript files, lazy-loaded chunks, and source maps (SKILL.md).
  • Boundary markers: None specified in the methodology.
  • Capability inventory: The skill methodology mentions external tools (katana, linkfinder, jsluice) but does not provide scripts, subprocess calls, or file-write capabilities within the skill itself.
  • Sanitization: Not addressed for the data being analyzed.
  • [SAFE]: No malicious instructions, sensitive file access, or network exfiltration patterns were detected in the instructions or metadata.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 11:01 PM
Security Audit — agent-trust-hub — recon-js-analysis