Direction before execution
A Quality Brief — visual direction, three defining words, constraints, references — must be explicitly approved before a single line of code is written.
A design-first AI workflow / v1.0 / MIT
Taste arrives last. Rework is the norm.
Taste leads. Code is the execution.
A Quality Brief — visual direction, three defining words, constraints, references — must be explicitly approved before a single line of code is written.
Scope freezes the moment a spec hits VALIDATED. Additions require a fresh RAY cycle. No silent creep, no moving targets.
ANALYZER writes a learning after every verdict. RAY reads the three most recent before the next spec — improvements compound.
Challenges the idea before it grows. Tiers every spec T1 / T2 / T3, locks scope, and reads the latest learnings first.
/rayWrites the Quality Brief, then implements through a 6-step Ralph Loop. No code without an approved direction.
/bobGrades every feature /20 — spec, design system, technical quality & CX — then writes the learnings that feed the next cycle.
/analyzerValidates that the problem is real before any brief is drafted. Kills bad ideas cheaply.
/eveRuns the deployment checklist and coordinates the hand-off so nothing slips between done and live.
/shipValidate the need before spending a single token on a solution.
→ Problem briefRight-size the work, define what's out of scope, freeze it.
→ Spec · VALIDATEDQuality Brief approved up front; atomic commits per step after.
→ Quality Brief + codeGraded out of 20 against the brief. The verdict gates release.
→ Score /20 + learningBefore BOB writes anything, the direction is set in writing — and approved. It's the difference between hoping the output has taste and guaranteeing it.
Read the spec on GitHub →Five minutes. Seven questions.
Generates STACK.md · CLAUDE.md · agent-system/ · the ANALYZER scoring framework