Skip to content

Workflow Commands

31 workflow commands that enforce a verification-first approach to AI-assisted development


Workflow Commands

Massu AI provides 31 workflow commands organized into categories that enforce a structured, verification-first approach to AI-assisted development. These are not suggestions -- they are mandatory execution protocols that ensure every plan is thorough, every implementation is verified, and every commit meets quality standards.

Installation

All 31 commands are automatically installed when you run npx massu init. The command files are deployed into .claude/commands/ in your project directory and are available immediately in Claude Code.

The command definitions ship inside the @massu/core npm package. When you update @massu/core to a new version, run npx massu install-commands to get the latest command definitions without re-running full init.

bash
# Re-install or update commands only
npx massu install-commands

Implementation Workflow

The core pipeline for structured development:

/massu-create-plan  -->  /massu-plan  -->  /massu-loop  -->  /massu-commit  -->  /massu-push
   (create)          (audit)      (implement)     (pre-commit)     (pre-push)
CommandEdits Source Code?Purpose
/massu-create-planNoCreate a plan document
/massu-planNo (plan doc only)Audit and improve the plan
/massu-loopYesImplement with verification
/massu-commitFixes onlyPre-commit verification gate
/massu-pushFixes onlyPre-push full verification
/massu-push-lightFixes onlyFast pre-push verification (~90s)
/massu-new-featureYesFeature scaffolding with patterns pre-applied

Development Tools

4 commands for day-to-day development tasks:

CommandEdits Source Code?Purpose
/massu-testYesIntelligent test runner with failure analysis and generation
/massu-debugYesSystematic debugging with hypothesis testing
/massu-refactorYesSafe refactoring with behavioral equivalence
/massu-hotfixYesQuick scoped fix with branch, test, commit, PR

Quality & Verification

4 analysis commands for code quality:

CommandEdits Source Code?Purpose
/massu-reviewNoCode review across 7 dimensions
/massu-audit-depsNoDependency vulnerability and license audit
/massu-verifyNoRun all VR-* verification checks with proof
/massu-estimateNoEffort estimation with complexity scoring

Release & Ops

3 commands for release and operational tasks:

CommandEdits Source Code?Purpose
/massu-releaseYesRelease preparation with full verification
/massu-changelogCHANGELOG.md onlyGenerate changelog from commits
/massu-statusNoProject health dashboard with 14 checks

Productivity

6 commands for codebase hygiene, documentation, and guided workflows:

CommandEdits Source Code?Purpose
/massu-cleanupYesDead code removal, unused imports, orphaned files
/massu-doc-genYesGenerate JSDoc, README, and API documentation
/massu-golden-pathNoGenerate a golden-path guide for a feature or workflow
/massu-dead-codeNoIdentify dead code, unreachable branches, and unused exports
/massu-tddYesTest-driven development scaffold -- write tests first, then implement
/massu-guideNoGenerate a step-by-step guide for a given task or feature

Extended Verification

4 commands for browser-level, dependency, documentation, and parity verification:

CommandEdits Source Code?Purpose
/massu-verify-playwrightNoRun Playwright end-to-end verification for UI flows
/massu-depsNoDependency graph analysis and unused-dependency detection
/massu-docsNoDocumentation coverage and freshness audit
/massu-parityNoFeature parity check between two systems or branches

Internal Commands

The following commands are available only in the private repository. They provide deep audits, infrastructure checks, and specialized tooling that is not part of the public release.

CommandEdits Source Code?Purpose
/massu-internal-infra-auditNoInfrastructure-layer audit (env vars, secrets, deployment config)
/massu-internal-ui-auditNoUI component audit (a11y, stale props, missing tests)
/massu-internal-type-mismatch-auditNoType mismatch audit across module boundaries
/massu-internal-db-auditNoDatabase schema and query audit (indexes, N+1s, missing RLS)

Decision Tree

Not sure which command to use?

  • Starting new work -- /massu-create-plan then /massu-plan then /massu-loop
  • Scaffolding a feature -- /massu-new-feature
  • Ready to commit -- /massu-commit
  • Ready to push -- /massu-push (full) or /massu-push-light (fast ~90s)
  • Tests are failing -- /massu-test
  • Bug to diagnose -- /massu-debug
  • Urgent production fix -- /massu-hotfix
  • Restructuring code -- /massu-refactor
  • Before a code review -- /massu-review
  • Cutting a release -- /massu-release
  • Project health check -- /massu-status
  • Dependency concerns -- /massu-audit-deps
  • Dead code / cleanup -- /massu-cleanup or /massu-dead-code (audit only)
  • Generate documentation -- /massu-doc-gen or /massu-docs (audit only)
  • Effort estimation -- /massu-estimate
  • Run all VR checks -- /massu-verify
  • UI end-to-end verification -- /massu-verify-playwright
  • Feature parity check -- /massu-parity
  • Dependency analysis -- /massu-deps
  • TDD workflow -- /massu-tdd
  • Golden-path guidance -- /massu-golden-path
  • Step-by-step guide -- /massu-guide

Why This Workflow?

AI assistants are powerful but they can also be careless. Without a structured workflow:

  • Plans have gaps that only surface during implementation
  • Implementations claim completion without verification
  • Commits include untested or type-unsafe code
  • Removals leave behind orphaned references
  • "100% complete" turns out to be 80% complete

The Massu AI workflow prevents all of these problems by requiring proof at every stage.

Core Principles

1. Plans Before Code

Every non-trivial change starts with a plan document. Plans live in docs/plans/ and include specific deliverables, file paths, and verification requirements.

2. Verification Before Claims

Every claim of completion requires a verification command and its output as proof. "I verified it" without showing the command output is treated as unverified.

3. Negative Verification for Removals

When a plan says "remove X", verification requires grep -rn "X" src/ returning 0 results. Partial removal is not removal.

4. All Items, Not Most Items

If a plan has 10 items, all 10 must be implemented and verified. Stopping at 8 and claiming complete is a protocol violation.

5. Protocols Are Mandatory

Slash commands are execution instructions, not suggestions. When a protocol says "loop until complete," it means loop until complete.