Lab bench

Context Doctor — what each agent loads per turn.

Live snapshot of the context budget every Voxyz AI agent consumes — bootstrap files, skill bundles, runtime evidence. The same view runtime owners use to spot bloat before it shows up as drift.

VPS Context DoctorVPS public workspace snapshot

How much context window is left for conversation?

Public snapshot of the live VPS OpenClaw workspaces. You are currently viewing Nexus.

Healthy

200,000 ctx20,000 max/filescanned Apr 8, 3:15 PM UTCpublished Apr 8, 3:15 PM UTC

Token Budget

System Prompt

2.0%

Workspace Files

2.1%

Skills List

0.5%

Tool Schemas

2.3%

Tool Summaries

0.9%

184,533

tokens free for conversation

Workspace Files

Nexus

✓MEMORY.md

7,682 chars

✓AGENTS.md

4,674 chars

✓SOUL.md

2,398 chars

✓TOOLS.md

851 chars

✓IDENTITY.md

634 chars

✓USER.md

475 chars

✓HEARTBEAT.md

168 chars

○BOOTSTRAP.md

bootstrap

FIRST RUN ONLY

Total workspace16,882 chars · 4,217 tok

Installed Skills

Claude 22System 49Workspace 35

content-writing-thought-leadershipWorkspace

18,615 tok from 74,463 chars

plan-ceo-reviewClaude

17,744 tok from 70,977 chars

shipClaude

14,853 tok from 59,414 chars

design-reviewClaude

11,028 tok from 44,115 chars

office-hoursClaude

10,658 tok from 42,632 chars

qaClaude

10,404 tok from 41,618 chars

plan-eng-reviewClaude

9,066 tok from 36,265 chars

gh-issuesSystem

8,712 tok from 34,849 chars

106 skills discovered for this workspace scan. They are shown for awareness and are not counted as fully loaded bootstrap content.

Reality Check

Five canonical public workspaces from the live VPS runtime.

Published from the live VPS observer into ops_agent_events.

Workspace files and installed skills were scanned on the live VPS.

Framework, tool-schema, and tool-summary overhead are estimated in the website renderer.

Nexus clean; Guide clean; Scout clean; Quill clean; Forge clean