skills/steipete/agent-scripts/npm/Gen Agent Trust Hub

npm

Warn

Audited by Gen Agent Trust Hub on Oct 18, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The script scripts/npm-auth-login.mjs dynamically loads the npm-profile library using require() on paths constructed at runtime. It searches through several global installation directories, including the result of npm root -g, /opt/homebrew/lib/node_modules, and /usr/local/lib/node_modules. This dynamic loading pattern from computed paths can be exploited to load malicious code if an attacker can influence the contents of these directories or the output of the npm configuration.
  • [INDIRECT_PROMPT_INJECTION]: The skill's scripts process package names and versions that may originate from untrusted external data provided via the agent's prompt, creating a potential injection surface.
  • Ingestion points: Package names and arguments provided to the shell scripts reserve-packages.sh, publish-package.sh, and npm-service.sh.
  • Boundary markers: The skill does not define specific delimiters or security warnings to distinguish external data from the script's operational logic.
  • Capability inventory: The skill can write files (such as package.json and .npmrc), execute shell commands via npm (which may trigger lifecycle scripts), and interact with the 1Password CLI (op).
  • Sanitization: While scripts use standard shell quoting (e.g., "$name"), they do not perform deep validation or sanitization of input strings before interpolating them into file templates or command arguments.
  • [COMMAND_EXECUTION]: The skill's core functionality relies on executing external binaries including npm, op (1Password CLI), tmux, node, and jq. These tools are used to manage authentication tokens and perform registry operations, representing a high-privilege execution surface.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 18, 2026, 02:02 PM