ln-781-build-verifier

Pass

Audited by Gen Agent Trust Hub on Mar 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes various system commands to build projects and manage dependencies.
  • Evidence: Triggers CLI tools including npm, yarn, pnpm, dotnet, pip, poetry, go, cargo, mvn, and gradle based on file markers in the project root.
  • [EXTERNAL_DOWNLOADS]: Fetches software components and configuration from external sources.
  • Evidence: Restores project dependencies by downloading packages from public registries such as npmjs.org, PyPI, NuGet, and Crates.io.
  • Evidence: The diagram.html file loads the Mermaid.js library from the JSDelivr CDN.
  • [REMOTE_CODE_EXECUTION]: Executes build-time scripts which may contain arbitrary code.
  • Evidence: Build processes for several supported languages (e.g., Node.js lifecycle scripts or Python's setup.py) involve the execution of code defined within the untrusted project files.
  • [PROMPT_INJECTION]: Vulnerable to indirect prompt injection through the analysis of project metadata.
  • Ingestion points: Scans the project root directory and reads marker files such as package.json, pyproject.toml, and Cargo.toml in SKILL.md.
  • Boundary markers: Absent. There are no delimiters or instructions to ignore embedded malicious content within the detected project files.
  • Capability inventory: Features extensive subprocess execution capabilities across multiple runtimes and compilers.
  • Sanitization: No sanitization or validation logic is defined for the content of the project markers before they influence command execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 8, 2026, 11:56 PM