git-commit-push-realtime

Pass

Audited by Gen Agent Trust Hub on Jul 29, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute various shell commands, primarily focused on Git operations (e.g., git status, git diff, git commit, git push) and repository state inspection.
  • [DATA_EXFILTRATION]: The core functionality involves pushing local code to a remote repository. The risk of accidental sensitive data exposure is addressed by instructions that forbid global staging (git add .) and mandate checks for secrets and environment files.
  • [REMOTE_CODE_EXECUTION]: Verification steps include running local project scripts like pnpm test and pnpm typecheck. While these are standard for development, they involve executing code present in the user's workspace.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from git log and git diff outputs.
  • Ingestion points: git diff, git log, and git status output analysis (SKILL.md).
  • Boundary markers: The skill requires an initial plan and diff reviews but does not define explicit delimiters to isolate git output from the agent's internal instructions.
  • Capability inventory: The agent possesses the capability to perform git commits, pushes, and execute arbitrary project tests (SKILL.md).
  • Sanitization: There is no mention of sanitizing or escaping the content retrieved from git commands before the agent processes it.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 29, 2026, 08:21 AM
Security Audit — agent-trust-hub — git-commit-push-realtime