npm-publish-tool

Pass

Audited by Gen Agent Trust Hub on Sep 2, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructions direct the agent to execute various shell commands to manage and release npm packages.
  • Evidence: The workflow in SKILL.md requires running bun install, bun run, bun pm pack, and bun publish to process the repository.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository's manifests and source code, which could potentially contain malicious instructions targeting the agent.
  • Ingestion points: Step 1 of the workflow in SKILL.md involves reading repository instructions, manifests like package.json, source files, and lockfiles.
  • Boundary markers: The skill does not provide specific delimiters or instructions to prevent the agent from obeying instructions embedded in the project files.
  • Capability inventory: The skill utilizes Bun for package management, script execution, and network operations to publish artifacts.
  • Sanitization: Step 5 includes a requirement for the agent to manually verify that secrets and local configurations are absent from the generated tarball.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 2, 2026, 06:59 PM
Security Audit — agent-trust-hub — npm-publish-tool