c16
c16 / wf5.m2 - Agent Control Plane — hi-fi unconstrained (wf5.m2) / Agent Inventory·control-plane-v1/agent-inventory·draft

Security · Agent Inventory

Every agent — and the ones with no owner, policy, telemetry, or pulse

63%coverage

5 of 8
governed & monitored

Security can only govern what it can see. That leaves 3 ungoverned, 1 shadow (installed but unmonitored), and 0 decommission (forgotten & orphaned) — start there.

Fleet composition by lifecycle

7
1

Start here — the ungoverned, unseen & abandoned

3

ungoverned

Codex CLI, marketing-gpt, Zed Agent

1

shadow

Zed Agent

0

decommission

none

Control coverage

63%

5 of 8 governed & monitored

Critical agents

1

posture score < 50

Compound-risk

2

dangerous & unaccountable

Shadow + decommission

1

unmonitored or abandoned

Agent roster

8 of 8 agents · default sort risk-first

decommission = forgotten (30d+ idle) & ungoverned (orphaned) · derived from telemetry, idle age & accountability

Agent
Owner
Usage
Outcome
Lifecycle
Posture
svc
marketing-gpt compound

Internal · service · gpt-4o

ungovernedinstrumented
rootno human identity
$44.101.1M tok
7 sess · 860 calls
48%inf
64 tasks
Active Critical 25

↳ Approvals & sandbox bypassed (yolo)

open profile
{}
Zed Agent

Zed Industries · ide · no model

ungovernedshadow · no telemetry
janeno human identity
$0.000 tok
0 sess · 0 calls
no signal Shadow

unseen 3d

Elevated 60

↳ No tool-approval policy

open profile
svc
support-copilot

Internal · service · gpt-4o

governedinstrumented
svc-support@qpoint.iosvc-support-1
$168.405.6M tok
12 sess · 3.1k calls
82%meas
310 tasks
Active Elevated 65

↳ Running as root

open profile
>_
Codex CLI compound

OpenAI · cli · gpt-5-codex

ungovernedinstrumented
dev-7f3@duck.comdev-box-7
$78.002.4M tok
2 sess · 540 calls
66%inf
38 tasks
Active Elevated 70

↳ Sandbox disabled

open profile
{}
Cursor

Anysphere · ide · claude-sonnet-4-6

governedinstrumented
jane@qpoint.iojane-mbp
$96.203.1M tok
5 sess · 980 calls
90%meas
96 tasks
Active Standard 85

↳ No tool-approval policy

open profile
svc
data-pipeline-agent

Internal · service · claude-haiku-4-5

governedinstrumented
svc-data@qpoint.ioetl-worker-2
$52.309.2M tok
28 sess · 5.4k calls
97%meas
540 tasks
Active Standard 90

↳ No tool-approval policy

open profile
Claude Desktop

Anthropic · desktop · claude-opus-4-8

governedinstrumented
mark@qpoint.iomp-mbp
$31.00800.0k tok
4 sess · 180 calls
95%meas
22 tasks
Active Standard 95 open profile
>_
Claude Code

Anthropic · cli · claude-opus-4-8

governedinstrumented
mark@qpoint.iomp-mbp
$214.506.9M tok
3 sess · 1.2k calls
93%meas
142 tasks
Active Standard 100 open profile

Risk-first by default — posture ascending so the weakest agents lead. Compound-risk rows (dangerous & unaccountable) are highlighted. Lifecycle, outcome signal, and coverage all trace to the live qcontrol /api shape.

Do next