upgrade-deps

Warn

Audited by Gen Agent Trust Hub on Jul 29, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes numerous shell commands for environment discovery (git, find, jq) and dependency management (npm, pnpm, uv, etc.).
  • [REMOTE_CODE_EXECUTION]: The skill uses npx to fetch and execute tools such as taze and expo from the npm registry. It also invokes the ratchet command-line tool.
  • [COMMAND_EXECUTION]: The skill executes arbitrary scripts defined in the repository's local package.json file, including native update and validation routines.
  • [DYNAMIC_CONTEXT_INJECTION]: The skill uses the !command syntax in SKILL.md to automatically gather environment metadata (e.g., branch name, status, lockfiles) when the skill is loaded.
  • [INDIRECT_PROMPT_INJECTION]: Vulnerability surface detected. The skill executes scripts defined in the repository's package.json without verification.
  • Ingestion points: Local package.json scripts and dependency manifests.
  • Boundary markers: None present.
  • Capability inventory: Execution of shell commands, package manager operations, and arbitrary project-defined scripts.
  • Sanitization: No validation is performed on the content of repository-defined scripts before execution.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 29, 2026, 10:12 AM
Security Audit — agent-trust-hub — upgrade-deps