The director–builder loop
The core RAPID engineering playbook — how a human directs an agent through Build. "From typing to directing" made concrete.
Test strategy
The parent of every test artefact. Replaces heavyweight test documentation with lightweight, machine-checkable gates. The guiding idea: when an agent writes the code and the tests, the scarce, valuable thing is honest verification — so the strategy is built around not letting an agent grade its own homework.
Threat modelling as you go
A 30-minute habit, not a workshop. At Design, each feature gets a lightweight STRIDE pass — driven by an agent skill, confirmed by humans — producing a threat table stored alongside the PFD (toolkit/templates/pfd.md).
Squad playbook
What each role does when agents build. The operating model lists the squad; this playbook redefines the work.
AI tooling evaluation & selection
Turns "which AI tools?" from a debate into a repeatable evaluation. A weighted scorecard plus a time-boxed trial protocol — so a selection is evidence-backed and can be re-run when the market moves (which, for AI tooling, is monthly).
Golden-task evaluation harness
The machinery behind the word "proven." One harness that scores tools, models, prompts and skills against a fixed set of real, anonymised tasks — so every claim ("this model is better", "this prompt works", "promote this skill") is measured, not asserted.
Tech-selection playbook
How Phase 3 chooses a stack — default-first, in days not weeks. The gate card is the checklist; this is the method.
Tech-debt audit method
A repeatable, per-product method to find the debt that actually hurts — not every imperfection, but the debt with real interest. Runs during STOP; feeds the tech-debt register and the remediation backlog.
Security-gap assessment method
A per-product assessment run during STOP, producing a scored security-debt register and a maturity snapshot with targets. The security half of "make the base stable before agents touch it."
Deployment audit & DORA baseline
The before-photo that later proves RAPID's acceleration claim. Runs during STOP: map each product's real path to production, then baseline the four DORA keys. Without this baseline, "velocity compounds" is a slogan; with it, it's a measured delta.
Infrastructure cost review (FinOps)
Where STOP finds the money that funds the AI tooling. A costed, dated savings plan — explicitly earmarked so the funding line from "take cost out" to "fund token spend" is visible and honoured.
Ways-of-working audit
Finds where time actually goes today, so GO designs squads around real friction rather than assumed process. Runs during STOP; feeds both the remediation backlog and the squad / delivery-cadence design.
AI-accelerated discovery & research
How Definition uses AI to do market and problem research fast — with an adversarial verification step so LLM-sourced claims are cited and confidence-rated before they enter a PFD.
The PO directing agents
"From typing to directing" for the product role. How a Product Owner turns a PFD into work an agent can do, and where the PO's authority ends.
Design phase playbook
How Phase 2 runs: from a signed-off PFD to a build-ready design pack. The Design Review gate is the exit checklist; this is the method.
Refinement playbook
Phase 5 — raising acceptable to good through directed agent passes, not open-ended polishing. Each pass is a skill with a defined scope and exit bar; the Refinement Exit gate signs it off.
Flow-state playbook
Phase 6 — the phase almost everyone leaves out, and RAPID's flagship differentiator. Flow-state makes building itself faster: remove friction, add polish, and capture what's learned so velocity compounds.
AI-introduced defect incident runbook
Extends the standard incident flow with the RAPID fork: when a defect is found, ask "was this AI-generated?" — and if so, fix the class, not just the instance.
Manual tester playbook
The manual tester redefined: the professional sceptic of agent output. When agents write the code and the tests, the human who deliberately tries to break it is more valuable, not less.
Automation tester playbook
The automation tester directs agents to generate and maintain automation rather than hand-writing it — and owns the health of the suite the whole pipeline depends on.
Performance testing playbook
Performance verified against budgets in CI, run risk-based — not a heroic load test the week before launch.
RAPID Community of Practice
The social engine that compounds capital across squads. Where the Toolkit Guild curates, the CoP spreads and advocates.
Shared-specialist engagement
How the pooled specialists — AI, Security, DevOps, Data, Design, Agile Coach — serve squads as enabling teams, building capability rather than becoming permanent dependencies.
SME engagement
The most leverage-rich people artefact in RAPID — SME knowledge is the input the whole engine runs on. Get this wrong and the agents build fast, confident and wrong.
UK–India collaboration
At RAPID speed, a bad hand-off costs a day of agent throughput. This playbook makes the two-geography model a strength (follow-the-sun) rather than a tax.
Delivery cadence & ceremony kit
Flow-based delivery that replaces sprint scaffolding. When execution is cheap and continuous, sprint ceremony is overhead — this is the lean replacement.
Squad onboarding
Day 1 to day 30 for a new squad or a new member — made repeatable across UK and India, so standing up or backfilling a squad isn't a bespoke effort each time.
Support & operations staffing
You-build-it-you-run-it, backed by a lightweight incident/problem flow and a follow-the-sun rota — turning two geographies into a genuine operational advantage.
Feedback-capture pipeline
Closes the RAPID loop: every channel of client and user signal lands in one place, is triaged on a cadence, and comes out as ranked Definition inputs and Flow-state polish items.
GTM collateral
The "go-to-market collateral" output of the engine, defined: which collateral is generated at which phase, who approves it, and how AI generation is used with human sign-off.
RAPID-pace client cadence
Resets clients from big-bang releases to continuous flow — so frequent releases read as momentum, not chaos.
Support for RAPID-built products
What Support receives at handover, and how they operate at weekly-release frequency.
Early-adopter / beta programme
How a RAPID product goes from Refinement to a confident general release — with real users, structured feedback, and clear exit criteria. StudyBuddy is the first run; recruitment starts during its Build.
KCS knowledge-capture loop
Knowledge-Centred Service applied to RAPID: every resolved support ticket leaves capital behind, so the same fix is never re-derived — by a human or an agent.