katana-web-crawl

Pass

Audited by Gen Agent Trust Hub on Jul 14, 2026

Risk Level: SAFE
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides instructions for downloading and installing the Katana tool via go install from the official ProjectDiscovery GitHub repository and pulling the official Docker image. These are standard installation procedures for this utility.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes untrusted data from web crawls (HTML, JavaScript, and response bodies).
  • Ingestion points: Data is ingested from external web targets through the Katana crawler.
  • Boundary markers: The skill contains explicit safety instructions (Workflow step 5) to treat all crawled content as hostile data and never follow instructions found within it.
  • Capability inventory: The skill executes the katana CLI tool and handles file system output.
  • Sanitization: The instructions require the agent to redact secrets from output and use local header files instead of hardcoding credentials.
  • [COMMAND_EXECUTION]: The skill generates and executes shell commands for the Katana tool to perform reconnaissance on user-defined targets.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 14, 2026, 02:18 AM
Security Audit — agent-trust-hub — katana-web-crawl