presubmit

Complete the required presubmit. Use before declaring a feature finished, after a large refactor or when asked.

⭐16stars🔀4forks📁View on GitHub🕐Updated Jan 3, 2026

When & Why to Use This Skill

The Presubmit Claude skill automates the execution of comprehensive test suites and static code analysis to ensure software quality. By integrating a robust battery of checks, it allows AI agents to verify code integrity, identify regressions, and maintain high development standards before finalizing features or completing large-scale refactors.

Use Cases

  • Feature Completion Validation: Automatically running a full suite of tests to ensure a new feature is fully functional and bug-free before marking the task as finished.
  • Refactoring Regression Testing: Executing extensive static checks and unit tests after significant code changes to guarantee that existing functionality remains intact.
  • Autonomous Debugging and Fixing: Enabling the AI agent to run presubmit scripts, interpret error logs, and iteratively apply fixes until all quality gates are passed.
  • Pre-deployment Quality Gate: Serving as a final local verification step to ensure the codebase meets all project requirements before it is pushed to a CI/CD pipeline.
namepresubmit
descriptionComplete the required presubmit. Use before declaring a feature finished, after a large refactor or when asked.
allowed-tools[

Presubmit

This skill provides the most extensive and robust battery of tests and static checks available.

Utility Scripts

  • presubmit.sh - Run the real presubmit script, capture output in a temp file

Instructions

The presubmit can take several minutes to complete. The script blocks until complete.

Important: Do not run the real presubmit directly, always use the utility script ${CLAUDE_PLUGIN_ROOT}/scripts/presubmit.sh.

Important: Ensure the cwd is the project root before running the presubmit.

  1. Run ${CLAUDE_PLUGIN_ROOT}/scripts/presubmit.sh.
  2. Check the output file. (It can take a few minutes for the presubmit to finish.)
  3. If the output contains errors unrelated to the present context, escalate to the user
  4. If the output contains errors related to your context, attempt to fix them
  5. If the error prove too hard to fix, summarize your findings, a more specific repro command and escalate to the user
  6. Run ${CLAUDE_PLUGIN_ROOT}/scripts/presubmit.sh again and repeat until it passes