npm-package-publishing

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted configuration files from the local repository to perform audits.
  • Ingestion points: package.json, lockfiles, .npmrc, .yarnrc.yml, and .github/workflows/* (SKILL.md).
  • Boundary markers: Absent; the skill does not define specific delimiters to isolate untrusted file content from its reasoning process.
  • Capability inventory: The agent can modify GitHub Actions workflows, update package manager configurations, and execute shell commands including npx tools and package builds (SKILL.md).
  • Sanitization: Absent; no explicit sanitization of ingested content is defined before the agent acts on the data.
  • [REMOTE_CODE_EXECUTION]: The skill utilizes npx to fetch and execute code from the npm registry for setup and auditing tasks.
  • Evidence: Recommends running npx @e18e/setup-publish, npx actions-up, and npx publint (SKILL.md).
  • [EXTERNAL_DOWNLOADS]: The skill references and downloads templates and tools from external repositories and services.
  • Evidence: Fetches setup templates from github.com/e18e/setup-publish and suggests the pnpm/setup action from the GitHub Marketplace (SKILL.md).
  • [COMMAND_EXECUTION]: The skill performs shell operations and package manager tasks to validate and configure the repository.
  • Evidence: Executes commands like npm ci --ignore-scripts, npm i -g npm, and zizmor (SKILL.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 12:01 PM
Security Audit — agent-trust-hub — npm-package-publishing