/mstack router
One sticky entry point that classifies intent, picks a playbook, copies steps into todos, and cites principles by name—without becoming a token bomb.
/mstack · playbooks · planning · 19 coding tools
AI doesn't needbetter code generation.It needs better engineering.
mstack gives coding agents a single /mstack router, evidence-gated playbooks, a grill-to-tickets planning pipeline, and session continuity—one CLI workflow across 19 AI coding tools, with Cursor as one option.
npm install -g @imisbahk/mstack/add-plugin mstack61 skills10 playbooks19 AI environmentsv0.6.0 current
01 Why mstack
The model isn't
the bottleneck.
Everyone is arguing about which coding model is better.
GPT.Claude.Gemini.Codex.
But a faster model still cannot rescue the wrong product, an undefined boundary, or a decision nobody wrote down.
The bottleneck is thinking: understanding users, choosing the right scope, designing the system, and verifying what happens when reality disagrees with the happy path.
mstack captures the workflow I use to take software from an idea to production—and routes agents through it with /mstack across every coding tool you use.
02 The system
One router.
Every engineering layer.
Not another project generator. A maintained set of playbooks, principles, planning skills, specialists, and safeguards that live with your repository—and reach 19 AI coding tools through one CLI.
Evidence-gated playbooks
10 playbooks with goal, finish condition, evidence, and constraints: bug fix, investigation, perf/hillclimb, feature, refactor, prototype, shipping, babysit PR/CI, autonomous run, and orchestrate.
Planning pipeline
Grill decisions before docs, keep a shared glossary and short ADRs, synthesize a spec, build a tracer-bullet ticket graph, implement the frontier in parallel, then run a two-axis review (standards vs spec).
Session continuity
Handoff, pickup, pause-safely, recall, and encode-lessons keep long agent runs resumable instead of inventing a new plan every session.
Principle skills
14 short, citeable principles drawn from Build Like This—so agents can name the rule that changed a decision instead of burying it in prose.
Profiles and model roles
Install lite or full. Bind code, judgment, and reviewPanel model roles (optional arena) so the right model does the right kind of work across 19 coding tools.
Nineteen-tool reach
One neutral engineering core for 19 verified environments—Claude Code, Cursor, Codex, Gemini CLI, Copilot, Windsurf, Warp, Amp, Zed, and more—plus opt-in specialist packs.
Specialist agents
19 focused roles with explicit responsibilities, boundaries, inputs, output contracts, and safe handoffs.
Skills, prompts, hooks
61 skills, 33 prompts, 4 local reviewable automations, and 14 reference templates that live with the repository.
Safe reconciliation
Recorded ownership, previews, approvals, backups, and recovery make updates inspectable and reversible—whether you install via CLI or the optional Cursor plugin.
19AI environments
19specialist agents
61engineering skills
10playbooks
14principles
15capability packs
03 Multi-agent engineering
One assistant is not
an engineering team.
Build Like This coordinates phase-gated specialists, running independent lanes in parallel where supported and named sequential passes elsewhere while one lead owns synthesis and shared decisions.
Idea
A hypothesis, not a specification.
Discovery lanes
Product and user researchers test the problem, users, needs, and alternatives in parallel where supported, or as named sequential passes elsewhere.
Decision gate
The Product Manager integrates evidence into scope, non-goals, and a measurable outcome.
Design lanes
Architecture, data, security, UX, and operations make bounded decisions against one product definition—after grilling unsettled choices into glossary and ADRs.
Delivery lanes
Backend, frontend, database, and test specialists work concurrently on tracer-bullet tickets once contracts stabilize.
Review lanes
Two-axis review checks standards and spec fit; code, security, accessibility, performance, and release evidence stay independent.
Production
One release owner integrates the evidence, deploys with authority, and feeds learning back through encode-lessons.
Each specialist has a responsibility, strict boundaries, required inputs, an output contract, and rules that prevent overlapping edits or recursive agent sprawl.
Parallel where supported and independent. Accountable at every gate.04 Build Like This
Code starts
halfway down.
The strongest engineering workflow begins before implementation and continues after deployment. Grill, document, ticket, implement, review—then hand off cleanly.
Idea
Start with a belief worth testing, not a stack worth using.
Users
Name the people whose behavior or outcome should change.
Needs
Separate observed problems from plausible assumptions.
Features
Choose the smallest behavior that can prove useful.
Product
Grill unsettled decisions, then make scope, non-goals, success, and evidence legible in product.md.
Architecture
Confirm the ledger, then define boundaries, contracts, data, security, and failure in architecture.md.
Spec & tickets
Synthesize a delivery spec and a blocking ticket graph before parallel implementation.
Implement
Ship the frontier of unblocked tickets; keep domain rules out of transports and UI.
Review & deploy
Two-axis review, then verify readiness, release progressively, observe, and recover safely.
Improve
Handoff, recall, and encode lessons—then return to the earliest phase affected by new evidence.
05 Why it works
Good software is
a chain of good decisions.
Thinking before typing
Fast code is useful only after the problem, constraints, and intended outcome are clear.
User before feature
Name the person and outcome before inventing surface area.
Contracts before clients
Stabilize permissions, errors, and shapes before UI and adapters race ahead.
Prove it works
Finish conditions need evidence—success paths, denial, failure, and recovery.
Smallest complete slice
Ship a thin vertical that is real, not a horizontal half of everything.
Never block on the human
Ask once with a default, then keep moving inside the granted authority.
Don't ask AI to build your app.
Give it an engineering system to work inside.
06 CLI install
One CLI.
Every coding agent.
Requires Node.js 20.11 or newer. Run init for project planning, then ai setup for agents, skills, prompts, and runtime guidance across 19 AI coding tools. Preview every repository change before applying it. Current release: v0.6.0.
npm install -g @imisbahk/mstack07 Optional · Cursor
Also available
as a Cursor plugin.
Prefer Cursor's plugin channel? /add-plugin mstack installs the same /mstack runtime. Use the plugin or mstack ai setup—not both for the same skills.
/add-plugin mstack