v1-hybrid-loop-spec · supabase_json
Limitless V1 Hybrid Thought-to-Front-Office + Content Loop
v1-hybrid-loop-spec artifact · for Limitless OS · status draft
open questions
- Who besides Seth should approve front-office drafts in V1: Betsy only, or another office lead?
- Which content formats matter first: short social posts, reels scripts, newsletter, FAQs, or all as draft options?
- Which front-office source should be connected first after MCP: PT Everywhere export, GoHighLevel/CM Hub, email, calls, or manual queue entry?
- What content claims require clinician review versus Seth-only approval?
Agent Handoff
Start Here
v1-hybrid-loop-spec artifact · for Limitless OS · status draft
Completion Evidence
No explicit evidence field yet. Require tests, screenshots, linked PRs, or reviewed outputs before marking complete.
Open Questions
- Who besides Seth should approve front-office drafts in V1: Betsy only, or another office lead?
- Which content formats matter first: short social posts, reels scripts, newsletter, FAQs, or all as draft options?
- Which front-office source should be connected first after MCP: PT Everywhere export, GoHighLevel/CM Hub, email, calls, or manual queue entry?
- What content claims require clinician review versus Seth-only approval?
Structured Payload
Machine-readable source fields
lanes
| id | title | agents | purpose | queue types | human owners |
|---|---|---|---|---|---|
| front_office_lane | Front-Office Action Lane | - intake_concierge - scheduler - missed_call_recovery - cancellation_recovery - patient_communication - human_approval_router - privacy_clinical_guardrail | Turn relevant thoughts into operational work for intake, scheduling, follow-up, and patient communication. | - new_lead - missed_call - unscheduled_eval - cancellation_recovery - referral_followup - patient_communication_draft - admin_payment_question - unclear_item_needing_human_triage | - Betsy/front office - Seth - clinician when clinical sensitivity appears |
| content_lane | Content Production Lane | - content_brain - social_post_drafter - newsletter_agent - patient_education_agent - human_approval_router - privacy_clinical_guardrail | Turn Seth’s thoughts into content and education drafts while preserving Seth’s voice and review control. | - content_idea - patient_education_topic - social_post_draft - reel_script - newsletter_topic - story_to_anonymize - referral_partner_education - campaign_angle - faq_blog_draft - seth_voice_operating_principle | - Seth - clinician/performance team for clinical claims - privacy/guardrail for patient stories |
| shared_command_guardrail_lane | Shared Command and Guardrail Lane | - chief_operating_brain - operating_memory_curator - decision_logger - human_approval_router - hipaa_privacy_guardrail - clinical_safety_reviewer - automation_boundary_agent | Make route/priority/approval decisions and prevent unsafe automation. | - daily_brief_item - decision_record - clarifying_question - policy_review - human_approval_request - memory_cleanup_request | - Seth - Betsy/front office - clinician/performance team |
title
Limitless V1 Hybrid Thought-to-Front-Office + Content Loop
source
atlas-codex
purpose
Design the first build slice for Limitless OS: Seth can think out loud in ChatGPT/Claude, and the system turns useful thoughts into operating memory, front-office action, content drafts, decisions, SOP/workflow candidates, or clarifying questions with human approval before external action.
client id
limitless
project id
07dee66f-8143-4f4d-9af9-b1ed413d2767
updated at
2026-05-09T20:06:02.591Z
v1 boundary
excluded
- No automatic patient-facing sends in V1.
- No autonomous schedule changes in V1.
- No diagnosis, plan-of-care changes, individualized clinical advice, or return-to-play clearance.
- No insurance coverage promises or collections decisions without human approval.
- No publishing content automatically.
included
- Capture thoughts from ChatGPT/Claude through the Limitless MCP.
- Classify thoughts into memory, decision, task, draft, SOP/workflow rule, or question.
- Route actionable thoughts to front-office and content lanes.
- Draft messages, scripts, content, and next-action recommendations.
- Route drafts to Seth, Betsy/front office, or clinician/performance review.
- Log outcomes back into memory for future improvement.
core promise
Seth captures messy thoughts once; Limitless OS routes the useful pieces into front-office execution, content production, operating memory, and approval queues.
project name
Limitless OS
artifact type
v1-hybrid-loop-spec
routing rules
- One thought may produce multiple outputs; for example, a runner insight can become memory, a social post draft, a patient education snippet, and a clinician-reviewed content note.
- Every external-facing output starts as a draft in V1.
- Clinical or patient-specific content routes through clinician review.
- PHI, patient stories, payment details, insurance details, and staff-sensitive notes route through guardrails before use.
- If confidence is low or owner is unclear, create a clarifying question instead of a task.
- If a thought describes repeatable behavior, create an SOP/workflow candidate even if no immediate task is created.
open questions
- Who besides Seth should approve front-office drafts in V1: Betsy only, or another office lead?
- Which content formats matter first: short social posts, reels scripts, newsletter, FAQs, or all as draft options?
- Which front-office source should be connected first after MCP: PT Everywhere export, GoHighLevel/CM Hub, email, calls, or manual queue entry?
- What content claims require clinician review versus Seth-only approval?
schema version
1
queue item shape
status values
- captured
- classified
- queued
- drafted
- awaiting_approval
- approved
- edited
- rejected
- sent_or_published
- deferred
- escalated
- logged
required fields
- task_id
- lane
- output_type
- role_topic_slug
- source_thought_ids
- summary
- sensitivity
- confidence
- draft_required
- approval_required
- human_owner
- due_timing
- status
routing pipeline
| name | step | description | data created |
|---|---|---|---|
| Capture | 1 | Thought Capture Agent stores the raw thought with source, timestamp, project, inferred metadata, and searchable embedding. | - brain_memory row - source=limitless-mcp - project=limitless-os - raw content - metadata - tags |
| Classify | 2 | Operating Memory Curator assigns one or more output types and tags. A thought may become multiple outputs. | - classification - topics - sensitivity - confidence - candidate outputs |
| Prioritize | 3 | Chief Operating Brain decides whether the thought is memory-only, daily-brief material, queue-worthy, clarification-needed, or blocked by policy. | - priority - route - brief inclusion - policy status |
| Route | 4 | The system creates front-office queue items, content queue items, decision records, SOP candidates, or clarification prompts. | - HCS queue item - role_topic_slug - source_refs - approval owner - due timing |
| Produce | 5 | The assigned specialist agent drafts the next useful artifact: message, call script, post, newsletter, SOP, summary, or question. | - draft output - reasoning summary - confidence - review notes |
| Approve | 6 | Human Approval Router sends the output to Seth, Betsy/front office, clinician/performance team, or privacy/clinical review based on sensitivity. | - approval request - review decision - edits - approval timestamp |
| Learn | 7 | Outcome is logged back to memory so future agent behavior improves. | - accepted/edited/rejected status - sent/published/scheduled outcome - conversion/result if known - new rule candidate |
success criteria
- Seth can capture real thoughts from ChatGPT/Claude and retrieve them later.
- The system correctly separates memory-only thoughts from actionable front-office and content opportunities.
- Front-office queue items are clear enough for Betsy/front office to act on.
- Content drafts preserve Seth’s voice and route sensitive claims through review.
- Human approval routing is obvious and conservative.
- Edited/rejected/approved outcomes improve future memory and rules.
thought output types
| type | becomes | example | description | human owner | primary agents |
|---|---|---|---|---|---|
| memory | - operating belief - content theme - clinical philosophy note - future retrieval context | Our runners often think cadence is the whole answer, but it is usually load tolerance. | Useful context that should be searchable and inform future work but does not require action yet. | Seth | - thought_capture_agent - operating_memory_curator |
| decision | - decision record - rationale - owner - downstream tasks | We should start with the front-office intake queue before clinical automation. | A stated choice, prioritization, or direction that should be preserved with rationale and owner. | Seth | - operating_memory_curator - decision_logger - chief_operating_brain |
| task | - queue item - owner - due timing - draft workflow - approval requirement | Betsy should not have to remember to follow up with unscheduled evals. | Something a person or agent should do, usually represented as an HCS queue item. | Betsy/front office or Seth | - chief_operating_brain - scheduler - intake_concierge - human_approval_router |
| draft | - social post draft - patient education snippet - email paragraph - front-office script | We need a better way to explain why waiting months for knee pain is a problem. | A thought whose next useful output is language: message, script, post, email, FAQ, education snippet, or newsletter section. | Seth for brand/content; front office or clinician for patient-specific messages | - content_brain - social_post_drafter - newsletter_agent - patient_education_agent - patient_communication |
| sop_workflow_rule | - workflow rule - checklist - automation candidate - future queue trigger - approval rule | Every missed call should get a text within 10 minutes if we cannot call back. | A repeatable operating rule, checklist, process, or automation candidate. | Seth/Betsy | - operating_memory_curator - workflow_improvement_agent - chief_operating_brain - human_approval_router |
| question | - clarifying question - decision prompt - policy question - blocked queue item | We need to be more aggressive with follow-ups. | A thought that is ambiguous, incomplete, risky, or needs owner intent before routing. | Seth or relevant domain owner | - operating_memory_curator - chief_operating_brain - privacy_clinical_guardrail |
human approval matrix
| owner | approves |
|---|---|
| Seth | - strategy - brand voice - content publish - business decisions - referral partner positioning - new automation classes |
| Betsy/front office | - scheduling messages - lead follow-up drafts - admin/payment drafts - front-office SOP changes |
| clinician/performance team | - clinical advice - patient-specific education - documentation drafts - return-to-sport or testing interpretation |
| privacy/guardrail review | - patient stories - PHI-sensitive use - HCS payload safety - automation boundary exceptions |
example transformations
| outputs | thought |
|---|---|
| - item: type: string, owner: string, result: string - item: type: string, owner: string, result: string - item: type: string, owner: string, result: string | People keep waiting too long to call us after knee pain starts. |
| - item: type: string, owner: string, result: string - item: type: string, owner: string, result: string | Betsy should not have to remember to follow up with unscheduled evals. |
| - item: type: string, owner: string, result: string - item: type: string, owner: string, result: string - item: type: string, owner: string, result: string | Our runners think cadence fixes everything, but it is usually load tolerance. |
| - item: type: string, owner: string, result: string | We need to be more aggressive with follow-ups. |