opensource-housekeeper

Fail

Audited by Gen Agent Trust Hub on Jul 29, 2026

Risk Level: HIGHCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to extract and execute arbitrary shell commands from untrusted README.md files under the 'Quick Start' section, creating a direct path for arbitrary command execution controlled by external project owners.
  • [COMMAND_EXECUTION]: Includes the use of sudo for sensitive operations, such as disabling the macOS firewall (sudo pfctl -d) and calibrating system time (sudo sntp), which constitutes a privilege escalation risk.
  • [REMOTE_CODE_EXECUTION]: Automatically triggers package managers (npm install, pip install, cargo build, etc.) on local project files. Malicious repositories can exploit lifecycle scripts (e.g., preinstall or postinstall) to execute arbitrary code during the installation process.
  • [EXTERNAL_DOWNLOADS]: Encourages the use of third-party GitHub mirror domains, such as gh-proxy.com, which have been flagged by security scanners for phishing. Using these unverified mirrors may expose the user to tampered or malicious repository content.
  • [EXTERNAL_DOWNLOADS]: Fetches configuration templates from GitHub's official repository (e.g., raw.githubusercontent.com/github/gitignore). While this source is a well-known service, it involves downloading and applying remote content to the local environment.
  • [PROMPT_INJECTION]: The skill processes local project files, documentation, and external search results without sanitization or boundary markers, establishing a significant surface for Indirect Prompt Injection attacks via untrusted project metadata.
Recommendations
  • HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/github/gitignore/main/Node.gitignore, https://raw.githubusercontent.com/github/gitignore/main/Python.gitignore - DO NOT USE without thorough review
  • AI detected serious security threats
  • Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
HIGH
Analyzed
Jul 29, 2026, 01:36 AM