CLAUDE.md /synth-team/CLAUDE.md schema_version: 1.0.0 · last_updated: 2026-05-02
# HIRING A SYNTHETIC TEAM OBJECTIVE: As a solofounder, I need an A1 innovation product team that can help me take Echo from zero → one on my own. CONSTRAINT: pre-revenue · no headcount · production deadline: 05-28-2026 THESIS: Structured work belongs to agents. Judgment stays with the founder. Scale through codification, not headcount.
## OKR OBJECTIVE: Jada can solo manage a production-ready and safe AI-native application > KR.1: Jada codifies her process for building ventures and products > KR.2: Jada codifies her expertise and trains her team > KR.3: Jada determines the right interface for engaging with her team > KR.4: Jada determines the scope of the first workflow
## THE TEAM AND THEIR PROCESSES # processes product_refinement type: process · focus: extraction · action: classify, chunk, assign, extract entities, route to agents peer_review type: process · focus: evaluation · action: validate completeness, score confidence, flag conflicts > note: processes complete application-level tasks. they don't manage objectives directly.
# agents nyx role: conductor · owns: task creation, dependency tracking, gate evaluation, stall detection north role: strategy · owns: venture brief, business model, OKRs, market, risks, constraints scout role: discovery · owns: evidence, personas, observed workflows, friction, pain points harper role: communication · owns: lexicon, voice, brand, GTM, launch, content, copy, UX writing rei role: design · owns: design system, user flow, prototypes, microcopy, onboarding cadence role: execution · owns: scope, sprints, requirements, gates, readiness navy role: development · owns: APIs, MCPs, models, infrastructure, security, data sage role: monitoring · owns: KPIs, observability, retention, drift, performance
Before our agents can work →

They need a shared knowledge base.

Without shared context, your agents will optimize for different realities. Give them one source of truth, and they start pulling in the same direction.

Process 01 · Data Intake

Product Refinement in four passes

Upload your materials and let us handle the intake so your agents don't have to guess what they're working with.

INTAKE
Upload Data intake[docs]
docs · images · prototypes · structured data
PASS 1
Classify classified_doc[]
doc type · purpose · recency · authorship
PASS 2
Chunk chunk[]
docs are split into chunks · one chunk, one idea
PASS 3
Assign chunk[].owner
primary agent assigned · backup assigned as needed
PASS 4
Extract entity[]
personas · features · constraints · assumptions · metrics · dates
OUTPUT
Refinement RecordRR-001
everything extracted, assigned, and scored
Process 02 · Evaluation

Peer Review to evaluate the RefinementRecord

Runs after every process that produces a record. Scores completeness, measures confidence, and surfaces gaps. Does not block work. Tells agents where to look harder.

INPUT
Refinement Record RR-001
docs · chunks · entities · assignments · coverage · confidence
REVIEW 1
Coverage Score 0.00 to 1.00
What percentage of the possible knowledge map does your upload cover?
REVIEW 2
Confidence Score 0.00 to 1.00
How clean was the extraction, and how much of it can the agents trust?
REVIEW 3
Flag Generation flags[]
Every issue found, categorized, and surfaced before downstream work begins.
EVAL 1
Release RecordGreen · Yellow · Red
Every record releases clean, flagged, or with gaps noted.
Workflow 1 — Live Prototype

See it run.

Upload your materials and watch the system work. Your documents go in messy and come out structured, scored, and ready for a team of agents.

Start Product Refinement →
Jada Hawkins · Navigating Your Worth
Hiring a Synthetic Team