pravidhi-commit-protocol
Pass
Audited by Gen Agent Trust Hub on Feb 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs extensive shell operations to manage Git branches, synchronize with remotes, and modify repository files like .gitignore.
- [COMMAND_EXECUTION]: The skill exhibits an indirect injection surface by extracting command strings from CI/CD configuration files (Ingestion points: .github/workflows/*.yml) and executing them locally (Capability: shell execution). There are no boundary markers or sanitization logic to vet these commands before execution.
- [REMOTE_CODE_EXECUTION]: The skill automates the installation and execution of various security and quality tools from public registries (npm, PyPI, Cargo, Go) at runtime. While these are standard developer tools, they involve executing code from external sources.
- [DATA_EXFILTRATION]: The skill is configured to push code and tags to remote repositories (git push) and create pull requests using CLI tools (gh, glab), which involves transmitting project data and metadata to external platforms.
- [COMMAND_EXECUTION]: The skill executes project-specific test suites and build scripts (e.g., npm test, pytest, npm run build), which involves running arbitrary code defined within the repository.
Audit Metadata