What problem does it solve?
AI agents often forget to run all required verification steps (builds, tests, architecture checks) after making code changes, leading to incomplete work that fails during review or deployment.
Core Features & Use Cases
- Automated Validation Loop: Guides you through running musts validate to get pending tasks, executing deterministic checks via musts run, and recording evidence for judgment checks via musts evidence until the report is clean.
- Parallel Task Processing: Supports dispatching independent judgment tasks to subagents for parallel execution, while sequencing tasks that share limited resources like simulators or databases.
- Noise Reduction: Includes guidance for adding .mustsignore files to exclude local artifacts from validation scope, preventing unnecessary re-validation of committed fixtures or generated files.
- Use Case: For a Rust repository with a MUSTS.yml defining cargo test and lint checks, use this skill to automatically run all pending validation tasks after code changes, record test logs as evidence, and confirm the repo is ready for PR submission.
Quick Start
Use the musts skill to run the full validation loop for your recent code changes, execute all pending deterministic checks, record evidence for any judgment tasks, and confirm the repository is ready for submission.