pnpm-audit-fix
Pass
Audited by Gen Agent Trust Hub on Aug 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes local shell commands (pnpm audit, pnpm update, pnpm install) and a bundled Node.js script to automate the detection and remediation of security vulnerabilities.
- [PROMPT_INJECTION]: The skill processes data from external sources (the pnpm registry) via command output, which presents a potential indirect prompt injection attack surface.
- Ingestion points: Vulnerability data is ingested from the output of pnpm audit --json, which is temporarily stored in /tmp/pnpm-audit.json and parsed by the collect-audit.mjs script.
- Boundary markers: A strict human-in-the-loop confirmation step is enforced in Step 4, requiring the user to approve a pre-fix report before any filesystem modifications occur.
- Capability inventory: The skill has the capability to modify package.json and project lockfiles through pnpm's update and install commands.
- Sanitization: The skill uses a dedicated Node.js helper script to parse the audit JSON and extract only specific required fields, which prevents the direct interpolation of raw, potentially malicious external strings into the agent's core instruction set.
Audit Metadata