skills/fikriaf/agentos/p5js/Gen Agent Trust Hub

p5js

Fail

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: HIGHCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
  • [DATA_EXPOSURE]: The Node.js script scripts/export-frames.js configures Puppeteer with --disable-web-security and --allow-file-access-from-files. These flags remove the browser's Same-Origin Policy and permit JavaScript to read arbitrary files from the local file system. In an environment where the agent generates the code to be executed, this creates a direct path for the exfiltration of sensitive files like ~/.ssh/id_rsa or .env files if a malicious prompt influences the code generation.
  • [DYNAMIC_EXECUTION]: The skill's primary workflow involves generating p5.js code (JavaScript) and executing it within a headless browser for rendering. This runtime execution of AI-generated code is a core feature but represents a medium to high risk finding when security sandboxing is explicitly disabled in the execution environment.
  • [COMMAND_EXECUTION]: The skill includes several shell scripts (scripts/render.sh, scripts/serve.sh, scripts/setup.sh) that execute system-level commands including node, ffmpeg, python3, and npx. The render.sh script passes user-supplied or agent-generated arguments (like file paths and resolutions) to these commands without robust sanitization.
  • [INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection.
  • Ingestion points: The agent creates HTML and JavaScript files based on user prompts and potentially external data sources mentioned in references/interaction.md (e.g., APIs, datasets).
  • Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the rendering pipeline.
  • Capability inventory: The skill possesses capabilities for network access (via browser), file system reading (via Puppeteer with security disabled), and shell command execution (scripts/render.sh).
  • Sanitization: There is no evidence of sanitization for the generated HTML/JS before it is processed by the headless renderer.
  • [EXTERNAL_DOWNLOADS]: The skill references and downloads p5.js, p5.sound.js, and other libraries from well-known and trusted CDNs including cdnjs.cloudflare.com, unpkg.com, and cdn.jsdelivr.net. These are documented as safe, neutral references to established services.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
May 19, 2026, 07:46 PM
Security Audit — agent-trust-hub — p5js