github-multi-repo

Warn

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill clones external repositories to a local temporary directory and executes 'npm update' and 'npm test' on the contents. This pattern can lead to arbitrary code execution if a repository contains malicious lifecycle scripts or tests.
  • [COMMAND_EXECUTION]: Extensive use of shell execution via 'Bash()' to interact with the GitHub CLI ('gh'), git, and the local filesystem for repository discovery and automation.
  • [EXTERNAL_DOWNLOADS]: Automates the fetching of remote code through 'gh repo clone' and 'gh api' calls to retrieve repository contents and configuration files.
  • [DYNAMIC_EXECUTION]: Programmatically generates and overwrites project configuration files ('package.json', 'config.json') and GitHub Action workflows ('integration.yml') across repositories.
  • [INDIRECT_PROMPT_INJECTION]:
  • Ingestion points: Fetches external data including repository lists, descriptions, and file contents from GitHub.
  • Boundary markers: None detected; external data is processed and used to determine subsequent commands and PR descriptions.
  • Capability inventory: Possesses significant capabilities including repository creation, file pushing, and execution of shell commands.
  • Sanitization: Uses 'jq' for processing structured JSON data, providing some structural validation, but does not explicitly sanitize natural language fields before use.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 4, 2026, 07:39 PM