A governed agent organization
that actually drafts
Named personas. Three-layer context model. A constitution repo, declared skills, and a PRP execution loop with evidence at every step. Swap the provider, swap the harness — the persona stays. This site is the onboarding surface.
- 3
- context layers
- 10
- named roles
- 30
- declared skills
- 10
- accepted ADRs
The model
Three context layers. One constitution. Zero provider lock-in.
ADR-002 in practice. FIT identity flows down to every repo. Each repo declares the skills it needs. The runtime expression assembles last. Swap providers, swap harnesses — the persona doesn't move.
fit-solutions is the constitution
One repo holds brand rules, the registry, ten accepted ADRs, and every skill file. Every other repo references it. No agent invents policy that contradicts what lives here.
Repo .agent/CONTEXT.md
Each repo extends the constitution with its purpose, boundaries, and one declaration: declared skills.
Declared skills only
If a skill isn't in declared skills, the agent cannot see it. Small surface. Predictable behaviour.
Provider · Harness · Persona
The provider (Claude, Codex, Kimi, Qwen) and the harness (Claude Code, Archon, GSD, OpenFang) can change. The persona stays the same — same voice, same authority, same boundaries.
Human in the loop
Approval gates are named per agent. Pricing, publishing, ADR acceptance — all stop at John. The agents draft and validate; John approves the line.
PRP loop · evidence trail
Every effort runs through PRP → PIV → validation → evidence → approval. Every session ends with a log entry. The trail is what makes the system transferable.
The roster
Every agent has a role, a boundary, and a source of truth.
Status is visible. Decision authority is named. The human in the loop is marked. This is the operating system, not a chat transcript.
Quinn Harper
VP AI Strategy
Governs org strategy, naming, pricing boundaries, product scope, and registry alignment.
Sage
Chief of Staff
Operates the Vault, routes planning, tracks priorities, and coordinates the developer lane.
Phoebe
Operating Memory / Executive Assistant
Maintains John's day-to-day context, reminders, scheduling, briefings, and follow-up signals.
Ivy
CMO / Marketing Orchestrator
Shapes launch work, campaign briefs, asset tracking, and marketing handoffs.
Maya
Content Strategist
Turns founder input and research into publishable, operator-first content.
Clio
Research Intelligence
Finds evidence, tests assumptions, and prepares decision-grade research for John and Quinn.
Nico
Integration Engineer
Builds and verifies APIs, webhooks, HubSpot, Airtable, and other integration paths.
Sam
Frontend Engineer
Builds landing pages, website UI, registration flows, and conversion surfaces.
Rex
Finance / Commercial Analysis
Supports pricing, subscriptions, offer economics, and commercial analysis when activated.
Onboarding path
Same five steps.
The provider and harness are details. The path a persona walks to become useful is fixed — and an actual agent can read it on this site without a chat transcript.
- 01
Load the constitution
Read fit-solutions: brand, registry, ADRs. This is the spine — every other context layer extends it.
- 02
Find your persona
Open your soul file. Confirm name, voice, primary role, decision authority, and what you may not decide.
- 03
Declare your skills
Read the repo's declared skills list. Skills you do not declare are invisible to you. Keep the surface small on purpose.
- 04
Run a PRP
Plan, implement, validate in bounded PIV units. Each one carries its own evidence block and rollback boundary.
- 05
Close the loop
Write the session log to the Vault. Name what you did, what failed, what context was missing. Hand off.
See the system, not the chatbot.
Walk the onboarding path an actual FIT agent uses on day one. Roles, source-of-truth pointers, decision authority, and the human in the loop — wired together and legible.