mapuamap
UserClaude Code plugin: m_* toolkit (m_plan plan->execute->verify, deploy/infra commands, Playwright full-test loop, m_code quality framework)
Categories
Indexed Skills (7)
m_code_init_project
Initialize or harden a project for AI-assisted development — CLAUDE.md, .claude/rules, architecture boundaries, seams for nondeterminism, deterministic check commands, first behavior tests, and CI readiness. Use when bootstrapping a new repo or making an existing one safe for repeated Claude-assisted change.
m_code_refactor
Refactor or restructure code safely in small, behavior-checked slices. Two behavior modes — preserve (legacy modernization, no behavior change) or may-change (architecture improvement) — each runnable as report-only or as a patch. Establishes a baseline and a test safety net before moving code, introduces seams before extraction, and justifies every new abstraction by concrete friction. Use for legacy cleanup, messy modules, monolith slices, dependency-direction fixes, ports/adapters, and AI-testability work.
m_code_rules_audit
Audit code against the rules that actually exist in this repo (CLAUDE.md, .claude/rules, configs, CI) plus the deterministic checks that can run safely — architecture boundaries, tests, typecheck, lint, build, security-sensitive changes. report-only by default. Use for PR/commit readiness and rule-compliance checks; complements /code-review, which is rule-agnostic.
m_plan_roll
Autonomous "full send" of the m_plan pipeline — plan at large scope and implement, no questions. Forces the full 9-artifact plan, decides every blocker itself (recording assumptions), then chains straight into m_plan_implement and drives to done + deploy + verify without pausing. Use when you explicitly want the whole architect → build → deploy → verify cycle run hands-off on a substantial task. Heavier and far less interactive than /m_plan — reach for it only when you truly want zero questions.
m_plan
Plan + Execute + Verify pipeline for a user-supplied task. Asks blocker/architectural questions upfront via AskUserQuestion, right-sizes the artifact set (4–9 docs depending on task class — tiny/small/medium/large), generates plans under .m_plan/<task-slug>/, then walks them step-by-step updating verification as it goes. /goal mode is opt-in. Use when user invokes /m_plan, says "spec this out and build it", "full plan + execute", asks to plan-and-execute a substantial change, or wants a complete architect → implement → deploy → verify cycle in one skill.
m_plan_implement
Execute and verify a plan previously generated by /m_plan, including deployment to the real declared environment rather than treating local checks as release proof. Loads existing artifacts from .m_plan/<task-slug>/, walks 05_step_plan.md, enforces 06_deploy_plan.md real-target deploy/smoke gates, updates 09_verification.md, then reports Done / Changed / Not done. /goal mode is opt-in. Use when user invokes /m_plan_implement, says "execute the plan", "run the m_plan", "implement the spec", "finish what was planned", or asks to resume a paused m_plan run.
m_verify
Feature-verification manager. Reads the .m_verify/ ledger of features that were implemented but not yet confirmed working, auto-runs every check AI can do on its own (build/test/lint/HTTP/CLI + Playwright/browser), hands the user a minimal checklist of only what a human must confirm, and spawns background repair agents for anything that fails — without blocking the conversation. Use when user invokes /m_verify, says "what do I need to test", "what's left to verify", "check the pending features", "verify what we built", or after a chunk of implementation work that needs sign-off.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.