appshot-web-videos
Pass
Audited by Gen Agent Trust Hub on May 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands to manage the video project, including
npm installfor dependency resolution,npm run devto start a local preview server, andnpx remotion renderto generate video files. It also utilizeslsofandkillfor process management. - [DATA_EXFILTRATION]: The skill requires broad read access to the target project's source code to extract features, routes, and brand elements for video generation. This broad file system access poses a data exposure risk if sensitive information is present in the codebase.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) by ingesting untrusted source code from the target project to determine video content and coding actions.
- Ingestion points: Target project source code (SKILL.md, Phase 1).
- Boundary markers: Absent; the agent is not instructed to isolate or ignore instructions within the scanned code.
- Capability inventory: File system writes and shell command execution (npm/npx).
- Sanitization: Absent; no validation or filtering of scanned content is specified.
- [EXTERNAL_DOWNLOADS]: The skill automates the installation of Node.js dependencies via
npm install, which involves downloading packages from external registries into the local project environment.
Audit Metadata