Fleet Ready Score

July 20, 2026 · View on GitHub

Public contract for npx @cobusgreyling/fleet-audit.

Mirrors Loop Ready so people who already trust loop-audit adopt fleet-audit without relearning the ritual.

Levels

Level	Score	Meaning	Safe autonomy
F0	0–39	Ad-hoc population	Humans only; catalog what exists
F1	40–64	Cataloged	Report + HITL inbox; no unattended prod writes
F2	65–84	Shared fleet	Budgets + kill switch + clone policy
F3	85–100	Enterprise fleet	Policy-as-code, compliance export, SLOs

Default CI gate for “we know our agents”: --min-score=40 (F1).
Default gate for “shared team fleet”: --min-score=65 (F2).

Score weights (target product UX)

Approximate contributions used by fleet-audit (base 10):

Signal	Points	Why
`FLEET.md` posture	+15	Written intent and sovereign control
`FLEET-STATE.md`	+12	Human-readable catalog
Registry present	+12	Machine list of agents
Manifests	+8–14	Per-agent identity + owner
Permissions model	+10	clone / run / edit clarity
Budget file	+10	Economics exist
Kill switch language	+8	Sovereign control
Accountability test	+10	The one-sentence standard
Patterns registry	+6	Intentional design
Audit workflow dogfood	+5	Continuous readiness
Schema valid / no shadows	penalties	Lies in the registry are worse than no registry
Inbox + audit runbooks	bonus via files	HITL path

Exact numbers may drift; levels and the accountability sentence must not.

Human output template

Fleet Ready: 48/100  ·  F1 cataloged

OK
  FLEET.md present
  Registry file found
  2 agent manifest(s)
  Fleet budget file present

WARN
  Missing inbox-runbook.md
  Kill switch not documented
  Accountability test not referenced

Suggested actions
  npx @cobusgreyling/fleet-init . --pattern shared-inbox-hitl
  Document kill switch + FLEET_PAUSE in FLEET.md
  Add the accountability sentence to FLEET.md

How to climb fast (solo, 30 minutes)

npx @cobusgreyling/fleet-init . --pattern team-agent-registry
Fill owners on each manifest
Add inbox pattern if any agent can write
Write kill switch paragraph
npx @cobusgreyling/fleet-audit . --suggest
Stop at F1 for a week — do not skip to unattended loops

Mapping from Loop Ready

Loop	Fleet expectation
L1 report-only	Register at F1
L2 assisted	F1 inbox + identity required
L3 unattended	F2 budgets + kill switch minimum

See maturity-model.md and accountability-test.md.