obscura-browser

Pass

Audited by Gen Agent Trust Hub on May 14, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses scripts/detect_obscura.py to search for obscura binaries in common directories such as Downloads, bin, and .local/bin. This verification step ensures the CLI is available before suggesting run commands to the user.
  • [EXTERNAL_DOWNLOADS]: Encourages the installation of development tools like hostc via npx and the acquisition of obscura binaries from public GitHub repositories. These are standard practices for developers to set up their local environment.
  • [DATA_EXFILTRATION]: Provides instructions for using hostc to tunnel local ports to public URLs. This workflow is intended to facilitate local testing by bypassing SSRF protections that block direct localhost access, making the local service accessible over the internet for the duration of the tunnel.
  • [PROMPT_INJECTION]: The skill is designed for web scraping and content extraction, which involves processing data from external URLs. This represents an inherent surface for indirect prompt injection, as malicious instructions or data embedded in target web pages could be ingested by the agent during scraping tasks.
Audit Metadata
Risk Level
SAFE
Analyzed
May 14, 2026, 07:57 AM
Security Audit — agent-trust-hub — obscura-browser