lightpanda-browser

Fail

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches pre-compiled binary files from a remote third-party source.
  • Evidence: SKILL.md contains curl commands targeting releases from the github.com/lightpanda-io/browser repository.
  • [REMOTE_CODE_EXECUTION]: Downloads and executes code from an external repository.
  • Evidence: The installation process involves downloading a binary file followed by chmod a+x and direct execution of the resulting file.
  • [COMMAND_EXECUTION]: Utilizes superuser privileges for system-level operations.
  • Evidence: The build-from-source instructions include sudo apt install for managing system dependencies.
  • [PROMPT_INJECTION]: Vulnerable to indirect prompt injection via untrusted web content ingestion.
  • Ingestion points: The skill fetches and renders content from arbitrary URLs using page.goto() in SKILL.md.
  • Boundary markers: Lacks specific delimiters or instructions to separate untrusted web content from system instructions.
  • Capability inventory: Enables network interaction, file writing, and binary execution.
  • Sanitization: No sanitization or filtering logic is detected for processing fetched HTML or text.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Mar 15, 2026, 11:49 PM