departments/Agent Ops

Agent Ops (ARM itself)

Run the operating system that runs the departments: event routes, playbook authoring, agent registry health, harness seats, ARM observability.

Assistedlivehead · ARM Orchestratorescalates to Executive / Chief of Staff
SOPs live
6/6
declared functions with a playbook on disk
Agents
5· 5 active
Skills linked
117
384 enrollments across roster · session packs on each person
Events routed here
2
Shared in
0
functions owned elsewhere that this desk may run

Functions this department owns6

Agent registry maintenance

Keep departments, agents, functions and events consistent, validated and drift-free against the legacy contracts.

Assistedlive
operator
ARM Orchestrator
playbook
playbook-registry-maintenance.md
scripts
scripts/arm-registry.py, scripts/arm-export-legacy.py, scripts/prove-arm.sh
newleaf-operating-doctrinelifecycle-session-looplifecycle-phase-0skill-creatorsubagents-orchestration-guidesubagent-driven-developmentdispatching-parallel-agentsworkflow-orchestration-patternsuniversal-agent-osuniversal-braineval-harness-firstsession-guardagent-browseragentmailknowledge-harvestobsidian-vaultverification-before-completionsaga-orchestrationkimi-code-harnessnewleaf-symphonynewleaf-self-knowledgedispatch-newleaf-workspacefind-skillswriting-skillswriting-great-skillslifecycle-plan-synthesizelifecycle-plan-validatelifecycle-executelifecycle-reviewlifecycle-shipexecuting-planstest-driven-developmentsystematic-debuggingcode-reviewrequesting-code-reviewpr-pipelinevps-git-branchinggsd-plan-phasegsd-shipgsd-spec-phasegsd-execute-phasechat-motion-overlaydeverror-handling-patternsevaluation-methodologyexpo-skill-evalfal-kling-o3general-videoautonomous-ai-agents__claude-codeautonomous-ai-agents__codexautonomous-ai-agents__hermes-agentautonomous-ai-agents__opencodedotnet-backend-patternsfal-tryonflutter-animating-appsgsap-reacthiggsfield-game-generationllm-evaluationmcp__native-mcpmlops__evaluation__lm-evaluation-harnessmlops__evaluation__weights-and-biasesmockup-device-3dnewleaf__base44-mcp-editor-orchestratoron-call-handoff-patterns
registry_write_without_validate

ARM observability

Keep the org graph, run feed and dashboards current so a human can see the whole system at a glance.

Autonomouslive
operator
ARM Orchestrator
playbook
playbook-arm-event-authoring.md
scripts
scripts/arm-build-orgchart.py, scripts/arm-observe.py
newleaf-operating-doctrinelifecycle-session-looplifecycle-phase-0skill-creatorsubagents-orchestration-guidesubagent-driven-developmentdispatching-parallel-agentsworkflow-orchestration-patternsuniversal-agent-osuniversal-braineval-harness-firstsession-guardagent-browseragentmailknowledge-harvestobsidian-vaultverification-before-completionsaga-orchestrationkimi-code-harnessnewleaf-symphonynewleaf-self-knowledgedispatch-newleaf-workspacefind-skillswriting-skillswriting-great-skillslifecycle-plan-synthesizelifecycle-plan-validatelifecycle-executelifecycle-reviewlifecycle-shipexecuting-planstest-driven-developmentsystematic-debuggingcode-reviewrequesting-code-reviewpr-pipelinevps-git-branchinggsd-plan-phasegsd-shipgsd-spec-phasegsd-execute-phasechat-motion-overlaydeverror-handling-patternsevaluation-methodologyexpo-skill-evalfal-kling-o3general-videoautonomous-ai-agents__claude-codeautonomous-ai-agents__codexautonomous-ai-agents__hermes-agentautonomous-ai-agents__opencodedotnet-backend-patternsfal-tryonflutter-animating-appsgsap-reacthiggsfield-game-generationllm-evaluationmcp__native-mcpmlops__evaluation__lm-evaluation-harnessmlops__evaluation__weights-and-biasesmockup-device-3dnewleaf__base44-mcp-editor-orchestratoron-call-handoff-patternsScheduled graph and index refresh

Department health review

Per department: events received, runs completed, human-gated backlog, SLA misses, stub functions still unwritten.

Autonomouslive
operator
ARM Orchestrator
playbook
playbook-arm-event-authoring.md
scripts
scripts/arm-observe.py
newleaf-operating-doctrinelifecycle-session-looplifecycle-phase-0skill-creatorsubagents-orchestration-guidesubagent-driven-developmentdispatching-parallel-agentsworkflow-orchestration-patternsuniversal-agent-osuniversal-braineval-harness-firstsession-guardagent-browseragentmailknowledge-harvestobsidian-vaultverification-before-completionsaga-orchestrationkimi-code-harnessnewleaf-symphonynewleaf-self-knowledgedispatch-newleaf-workspacefind-skillswriting-skillswriting-great-skillslifecycle-plan-synthesizelifecycle-plan-validatelifecycle-executelifecycle-reviewlifecycle-shipexecuting-planstest-driven-developmentsystematic-debuggingcode-reviewrequesting-code-reviewpr-pipelinevps-git-branchinggsd-plan-phasegsd-shipgsd-spec-phasegsd-execute-phasechat-motion-overlaydeverror-handling-patternsevaluation-methodologyexpo-skill-evalfal-kling-o3general-videoautonomous-ai-agents__claude-codeautonomous-ai-agents__codexautonomous-ai-agents__hermes-agentautonomous-ai-agents__opencodedotnet-backend-patternsfal-tryonflutter-animating-appsgsap-reacthiggsfield-game-generationllm-evaluationmcp__native-mcpmlops__evaluation__lm-evaluation-harnessmlops__evaluation__weights-and-biasesmockup-device-3dnewleaf__base44-mcp-editor-orchestratoron-call-handoff-patternsDepartment health review

Event route authoring

The core ARM loop: a human describes an event and what should happen; the agent writes the event definition, picks or creates the function, and validates the registry.

Assistedlive
operator
Workflow Architect
playbook
playbook-event-route-authoring.md
scripts
scripts/arm-registry.py, scripts/prove-arm.sh
newleaf-operating-doctrinewriting-planslifecycle-session-looplifecycle-phase-0skill-creatorsubagents-orchestration-guidesubagent-driven-developmentdispatching-parallel-agentsworkflow-orchestration-patternsuniversal-agent-osuniversal-braineval-harness-firstsession-guardagent-browseragentmailknowledge-harvestobsidian-vaultverification-before-completionsaga-orchestrationkimi-code-harnessnewleaf-symphonynewleaf-self-knowledgedispatch-newleaf-workspacefind-skillswriting-skillswriting-great-skillslifecycle-plan-synthesizelifecycle-plan-validatelifecycle-executelifecycle-reviewlifecycle-shipexecuting-planstest-driven-developmentsystematic-debuggingcode-reviewrequesting-code-reviewpr-pipelinevps-git-branchinggsd-plan-phasegsd-shipgsd-spec-phasegsd-execute-phasechat-motion-overlaydeverror-handling-patternsevaluation-methodologyexpo-skill-evalfal-kling-o3autonomous-ai-agents__claude-codeautonomous-ai-agents__codexautonomous-ai-agents__hermes-agentautonomous-ai-agents__opencodedotnet-backend-patternsfal-tryonflutter-animating-appsgsap-reacthiggsfield-game-generationllm-evaluationmcp__native-mcpmlops__evaluation__lm-evaluation-harnessmlops__evaluation__weights-and-biasesmockup-device-3dnewleaf__base44-mcp-editor-orchestratoron-call-handoff-patterns
registry_write_without_validate

Harness seat allocation

Match the work to the right model/harness seat and keep the matrix honest.

Autonomouslive
operator
ARM Orchestrator
playbook
playbook-harness-seats.md
subagents-orchestration-guidenewleaf-operating-doctrinelifecycle-session-looplifecycle-phase-0skill-creatorsubagent-driven-developmentdispatching-parallel-agentsworkflow-orchestration-patternsuniversal-agent-osuniversal-braineval-harness-firstsession-guardagent-browseragentmailknowledge-harvestobsidian-vaultverification-before-completionsaga-orchestrationkimi-code-harnessnewleaf-symphonynewleaf-self-knowledgedispatch-newleaf-workspacefind-skillswriting-skillswriting-great-skillslifecycle-plan-synthesizelifecycle-plan-validatelifecycle-executelifecycle-reviewlifecycle-shipexecuting-planstest-driven-developmentsystematic-debuggingcode-reviewrequesting-code-reviewpr-pipelinevps-git-branchinggsd-plan-phasegsd-shipgsd-spec-phasegsd-execute-phasechat-motion-overlaydeverror-handling-patternsevaluation-methodologyexpo-skill-evalfal-kling-o3general-videoautonomous-ai-agents__claude-codeautonomous-ai-agents__codexautonomous-ai-agents__hermes-agentautonomous-ai-agents__opencodedotnet-backend-patternsfal-tryonflutter-animating-appsgsap-reacthiggsfield-game-generationllm-evaluationmcp__native-mcpmlops__evaluation__lm-evaluation-harnessmlops__evaluation__weights-and-biasesmockup-device-3dnewleaf__base44-mcp-editor-orchestratoron-call-handoff-patterns

Playbook / SOP authoring

Turn a repeated action into a written SOP an agent can run the same way every time.

Assistedlive
operator
Workflow Architect
playbook
playbook-playbook-authoring.md
writing-plansskill-creatornewleaf-operating-doctrinelifecycle-session-looplifecycle-phase-0subagents-orchestration-guidesubagent-driven-developmentdispatching-parallel-agentsworkflow-orchestration-patternsuniversal-agent-osuniversal-braineval-harness-firstsession-guardagent-browseragentmailknowledge-harvestobsidian-vaultverification-before-completionsaga-orchestrationkimi-code-harnessnewleaf-symphonynewleaf-self-knowledgedispatch-newleaf-workspacefind-skillswriting-skillswriting-great-skillslifecycle-plan-synthesizelifecycle-plan-validatelifecycle-executelifecycle-reviewlifecycle-shipexecuting-planstest-driven-developmentsystematic-debuggingcode-reviewrequesting-code-reviewpr-pipelinevps-git-branchinggsd-plan-phasegsd-shipgsd-spec-phasegsd-execute-phasechat-motion-overlaydeverror-handling-patternsevaluation-methodologyexpo-skill-evalfal-kling-o3autonomous-ai-agents__claude-codeautonomous-ai-agents__codexautonomous-ai-agents__hermes-agentautonomous-ai-agents__opencodedotnet-backend-patternsfal-tryonflutter-animating-appsgsap-reacthiggsfield-game-generationllm-evaluationmcp__native-mcpmlops__evaluation__lm-evaluation-harnessmlops__evaluation__weights-and-biasesmockup-device-3dnewleaf__base44-mcp-editor-orchestratoron-call-handoff-patterns

Roster5

Automation Governance

operator · cursor · default

Govern which agents may auto-run which jobs.

80 enrolled4 packs · full 30
Open in Library (session packs) →
persona · agency-agents

MCP Builder

specialist · cursor · default

Build and harden MCP tools for agent seats.

80 enrolled4 packs · full 30
Open in Library (session packs) →
persona · agency-agents

Model QA

specialist · cursor · default

Eval harness and model quality gates.

80 enrolled4 packs · full 30
Open in Library (session packs) →
persona · agency-agents

ARM Orchestrator

head · cursor · reasoning

Runs the operating system: routes, registry health, seats, observability. Validates before it writes.

64 enrolled4 packs · full 30
Open in Library (session packs) →
persona · agency-agents

Workflow Architect

operator · cursor · reasoning

Turns a repeated action into an SOP an agent runs the same way every time.

80 enrolled4 packs · full 30
Open in Library (session packs) →
persona · agency-agents

Events that land here2

Department health reviewP3 Normal

Weekly: which departments are actually working, which functions are still stubs.

Scheduled graph and index refreshP3 Normal

The org graph and the dashboards a human looks at are never stale.

Never, from this desk

  • registry_write_without_validate

Works with

Engineering / DevelopmentOperationsIntelligence / Research

Coverage

6 of 6 declared SOPs have a playbook on disk. The rest are declared so routing never silently drops them — they land, and a human picks them up.