github-use
Pass
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes various git and GitHub CLI (gh) commands in the Windows PowerShell environment. It also instructs the agent to run a local script (./activate) to set up its environment.
- [DATA_EXFILTRATION]: The skill provides automated workflows for pushing local code to remote repositories (git push) and creating issues on GitHub (gh issue create). This capability could be misused to transmit sensitive local data to external platforms.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its automated data processing workflows.
- Ingestion points: The agent reads and processes repository file contents using tools like codebase_search and read_file during the 'Research changes' step. It also accepts untrusted user input for branch names and issue content.
- Boundary markers: While it uses double quotes for command arguments, it lacks specific boundary markers or instructions to disregard potential commands embedded within the files it analyzes.
- Capability inventory: The skill grants the agent the authority to execute shell commands and modify remote repository states without user confirmation ('Run all git commands without asking permission').
- Sanitization: It performs minimal sanitization by validating branch names and avoiding backticks in commit messages, but it does not sanitize the content being committed or included in issue bodies.
Audit Metadata