skills/cuioss/plan-marshall/build-npm/Gen Agent Trust Hub

build-npm

Pass

Audited by Gen Agent Trust Hub on Jul 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes npm and npx commands using a structured dispatcher (npm.py) and an execution factory that handles parameter passing and environment configuration.
  • [EXTERNAL_DOWNLOADS]: As a build system integration, the skill facilitates the download of third-party packages from the npm registry during npm install or npx execution.
  • [PROMPT_INJECTION]: The skill processes untrusted external data, including build logs (Jest, ESLint, TAP) and project descriptors (package.json, pnpm-workspace.yaml). This presents an indirect prompt injection surface where malicious content in a log file or package name could attempt to influence the agent.
  • Ingestion points: Project metadata files (package.json, pnpm-workspace.yaml) and build/test log output captured during execution.
  • Boundary markers: No specific delimiters or "ignore instructions" warnings are added to the parsed output before it is returned to the agent.
  • Capability inventory: The skill possesses the ability to execute shell commands (npm, npx) and read filesystem contents.
  • Sanitization: Log parsing is performed using specific regular expression patterns, which isolates extracted data fields from the raw log stream, though the extracted text (e.g., test failure messages) is not sanitized for natural language instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 11, 2026, 07:12 AM
Security Audit — agent-trust-hub — build-npm