git-cp
Warn
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- COMMAND_EXECUTION (LOW): The skill executes git commands (stage, commit, push) to perform its functions, which requires shell access.
- DATA_EXFILTRATION (MEDIUM): The push operation facilitates the transfer of local repository data to a remote server, which is an inherent exfiltration vector.
- Indirect Prompt Injection (Category 8) (MEDIUM): The agent reads untrusted local code changes to generate commit messages and plan tasks. * Ingestion points: Local branch code and file changes. * Boundary markers: None present. * Capability inventory: Git command execution and network operations (push). * Sanitization: None identified; the agent processes file content to 'create a meaningful commit'.
Audit Metadata