winning-web-speed-hackathon

Pass

Audited by Gen Agent Trust Hub on Mar 20, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill suggests cloning the official scoring tool from a repository at https://github.com/CyberAgentHack/web-speed-hackathon-scoring-tool and installing dependencies via pnpm install.
  • [COMMAND_EXECUTION]: Includes a utility script scripts/analyze-bundle.sh that executes shell commands (find, du, sort, head, numfmt, wc) to analyze the local file system within the build directory.
  • [COMMAND_EXECUTION]: Instructions explicitly call for executing various CLI tools and build commands, including npx lighthouse, npx playwright test, npx webpack-bundle-analyzer, npx glyphhanger, and image processing tools like magick and ffmpeg.
  • [PROMPT_INJECTION]: Indirect surface detected in scripts/analyze-bundle.sh which processes file names and extensions from the build directory without explicit sanitization.
  • Ingestion points: Reads directory content from the dist folder (or user-specified directory).
  • Boundary markers: None present; the script assumes standard file names.
  • Capability inventory: File system enumeration and metadata retrieval via find and du in scripts/analyze-bundle.sh.
  • Sanitization: Absent; shell commands are executed directly on found file paths.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 20, 2026, 01:53 AM
Security Audit — agent-trust-hub — winning-web-speed-hackathon