$ cat ./ASDLC.md --long

ASDLC: the Agentic Software Development Lifecycle

A structured method for shipping agent systems. Five stages, each with an exit gate, so agentic pace never outruns engineering rigour.

$ bh run --lifecycle

Five stages, five gates

[1] SPECIFY

Pin the problem

Define the job, its inputs and outputs, success criteria, and failure modes before a single token is spent.

[2] ARCHITECT

Design the system

Map boundaries, data flows, failure modes, and security posture before a line is built.

[3] BUILD

Agents + engineers

Build at agentic pace with human judgement on every change.

[4] VERIFY

Prove it works

Test, eval, and audit as first-class work - not an afterthought at handover.

[5] SHIP

Run in production

Deploy with observability, then hand over with your team trained to operate it.

# Skipping steps is how impressive demos become unmaintainable liabilities. We have audited enough vibecoded codebases to know.

$ ls ./asdlc --long

Inside each phase

1-specify.md
[1] SPECIFY

Pin the problem

Define the job, its inputs and outputs, success criteria, and failure modes before a single token is spent.

agents draft and stress-test requirements; a human signs off the spec. Clarity first - agents amplify clarity, and chaos.

  • exit gate: a specification the whole system can be tested against
2-architect.md
[2] ARCHITECT

Design the system

Map boundaries, data flows, failure modes, and security posture before a line is built.

agents generate and compare architecture options; a human picks the seams and approves the guardrails.

  • exit gate: an architecture with defined seams, tools, and guardrails
3-build.md
[3] BUILD

Agents + engineers

Build at agentic pace with human judgement on every change.

agents scaffold, implement, and iterate; humans review every change and every decision stays traceable.

  • exit gate: a working system with a traceable, versioned history
4-verify.md
[4] VERIFY

Prove it works

Test, eval, and audit as first-class work - not an afterthought at handover.

agents generate evals and edge cases; humans run red-teaming and audits.

  • exit gate: evidence the system is safe and correct before it ships
5-ship.md
[5] SHIP

Run in production

Deploy with observability, then hand over with your team trained to operate it.

agents monitor, flag, and draft runbooks; your team owns the system without us on speed dial.

  • exit gate: a production system your team operates confidently
$ diff ./asdlc ./vibe-coding

ASDLC vs vibe coding

asdlc-vs-vibe.log
 ASDLCvibe coding
structurefive explicit stages, each with an exit gatenone - prompt, paste, pray
testingevals, tests, and audits as first-class workrarely - "it worked on my machine"
reviewhuman review on every change, traceable decisionsnone or superficial
handoverdocumented, observed, your team operates itno one is sure how it works
failure modepredictable delivery, an operated assetimpressive demo becomes an unmaintainable liability

$ book --type discovery --duration 30m --cost $0

Thirty minutes.
Zero dollars. Real answers.

Bring the problem, the half-built prototype, or just the hunch that agents could change how you work. We’ll tell you straight what’s worth building — and what isn’t.

Book a free 30-min discovery call->

# no deck, no pressure, no obligation. exit code 0 either way.