node-deps
Warn
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [Unverifiable Dependencies & Remote Code Execution] (MEDIUM): The skill prompts for the installation of the 'taze' package and executes a local shell script (~/.claude/skills/node-deps/scripts/run-taze.sh) whose source is not verified.
- [Dynamic Execution] (MEDIUM): The skill assembles shell commands at runtime by concatenating package names found in the project's dependency list (e.g., in Step 6). This lacks sanitization, which could allow a malicious package name in package.json to execute arbitrary code via command injection.
- [Indirect Prompt Injection] (LOW): 1. Ingestion points: External data from taze CLI output parsed in SKILL.md. 2. Boundary markers: None present to distinguish data from instructions during shell interpolation. 3. Capability inventory: Subprocess shell execution for taze, npm, yarn, and pnpm. 4. Sanitization: None identified for package names used in command construction.
Audit Metadata