Skills
skills/everyinc/compound-engineering-plugin/feature-video/Gen Agent Trust Hub

feature-video

Pass

Audited by Gen Agent Trust Hub on Mar 24, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes Pull Request metadata (title, body) to automatically plan and execute browser interactions.\n
  • Ingestion points: PR details retrieved via gh pr view in the gather_context section of SKILL.md.\n
  • Boundary markers: Absent; there are no clear delimiters or instructions to separate PR content from the agent's logic.\n
  • Capability inventory: Includes browser automation via agent-browser, repository modification via gh pr edit, and network operations via rclone.\n
  • Sanitization: No sanitization of the PR body or title is performed before the planning stage.\n- [COMMAND_EXECUTION]: The skill uses several system commands including gh, ffmpeg, rclone, and agent-browser. These are used to interact with the repository, process video files, and manage cloud storage, which are consistent with the skill's primary function.\n- [EXTERNAL_DOWNLOADS]: The skill requests the installation of the agent-browser package from the NPM registry and its subsequent internal dependencies via agent-browser install.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 24, 2026, 12:12 AM