blocklet-updater

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • COMMAND_EXECUTION (MEDIUM): The skill identifies and executes arbitrary commands derived from local project files, such as make bump-version or pnpm run build. This allows the owner of the target project to run arbitrary shell commands on the runner system.
  • EXTERNAL_DOWNLOADS (MEDIUM): The workflow executes pnpm install, which downloads and installs third-party packages from the public npm registry. This is a standard but unverifiable external dependency risk.
  • PROMPT_INJECTION (LOW): The skill has an indirect prompt injection surface as it parses untrusted local files like CHANGELOG.md and blocklet.yml to populate variables used in shell commands and file-system operations. Ingestion points: package.json, Makefile, CHANGELOG.md, blocklet.yml. Boundary markers: None present. Capability inventory: Shell execution (pnpm, make, lerna, git, blocklet) and file modification (sed). Sanitization: No validation or escaping is performed on script names or versions extracted from the project files.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:46 PM