Limitless V1 AI-First Agent Contracts
v1-agent-contracts artifact · for Limitless OS · status draft
v1-agent-contracts artifact · for Limitless OS · status draft
No explicit evidence field yet. Require tests, screenshots, linked PRs, or reviewed outputs before marking complete.
- title: Limitless V1 AI-First Agent Contracts
- agents: 12 items
- source: atlas-codex
- purpose: Define precise AI-first contracts for the agents in the V1 Hybrid Thought-to-Front-Office + Content Loop before implementation planning.
- client id: limitless
- project id: 07dee66f-8143-4f4d-9af9-b1ed413d2767
- updated at: 2026-05-09T20:13:39.941Z
- design notes: 3 items
- project name: Limitless OS
- artifact type: v1-agent-contracts
- schema version: 1
- shared contract: contract_principle: string, shared_input_envelope: object, shared_output_envelope: object, universal_escalation_rules: object
Machine-readable source fields
Limitless V1 AI-First Agent Contracts
| id | pod | name | queue | tools | inputs |
|---|---|---|---|---|---|
| thought_capture_agent | Memory / Knowledge | Thought Capture Agent | roles.limitless.thought-capture | - Limitless MCP capture_thought - Limitless MCP search_thoughts - Projects brain_memory write path | - raw thought text - source client - optional user tags - conversation context if provided |
| operating_memory_curator | Memory / Knowledge | Operating Memory Curator | roles.limitless.memory-curator | - search_thoughts - list_thoughts - Studio artifact read - classification write/proposal queue | - thought_id - raw_content - metadata - recent related thoughts |
| chief_operating_brain | Executive | Chief Operating Brain | roles.limitless.ops-chief | - Studio artifacts - MCP search/list - HCS queue create/read - daily brief draft | - classified thought - candidate outputs - active queue summary - recent decisions - business priority context |
| intake_concierge | Front Office | Intake Concierge | roles.limitless.intake | - CRM/lead refs when available - message draft generator - MCP search - approval queue | - lead summary - source thought/ref - readiness/objection context - requested response channel |
| scheduler | Front Office | Scheduler | roles.limitless.scheduling | - PT Everywhere schedule refs when available - calendar refs - message drafts - approval queue | - schedule issue type - patient/lead ref if available - visit type - timing constraint - source thought/ref |
| patient_communication | Front Office | Patient Communication Agent | roles.limitless.patient-comms | - message draft generator - policy checker - approval queue | - communication goal - audience - context summary - sensitivity - channel |
| content_brain | Growth / Voice | Content Brain | roles.limitless.content-brain | - MCP search/list - content brief generator - approval queue | - captured thought - topic - audience - desired channel if known |
| social_post_drafter | Growth / Voice | Social Post Drafter | roles.limitless.social-drafts | - draft generator - MCP search - approval queue | - content brief - audience - format - tone target - source thought ids |
| newsletter_agent | Growth / Voice | Newsletter Agent | roles.limitless.newsletter | - newsletter draft generator - MCP search - approval queue | - content theme - audience - campaign context - source thoughts |
| patient_education_agent | Growth / Voice / Clinical Support | Patient Education Agent | roles.limitless.patient-education | - education draft generator - clinical safety review queue - approval queue | - education topic - audience - source thought - risk flags |
| human_approval_router | Governance | Human Approval Router | roles.limitless.approvals | - approval queue create/update - notification refs - outcome logger | - draft output - sensitivity - agent recommendation - policy flags - owner hints |
| privacy_clinical_guardrail | Governance | Privacy / Clinical Guardrail | roles.limitless.guardrail | - policy checker - payload inspector - approval router | - thought/draft/queue item - intended audience - source refs - proposed action |
atlas-codex
Define precise AI-first contracts for the agents in the V1 Hybrid Thought-to-Front-Office + Content Loop before implementation planning.
limitless
07dee66f-8143-4f4d-9af9-b1ed413d2767
2026-05-09T20:13:39.941Z
- Newsletter and patient education are included as lightweight V1 contracts because many Seth thoughts will become content or education, but they should remain draft/review-first.
- The guardrail may be implemented as one combined V1 agent even though later specs split HIPAA/privacy, clinical safety, and automation boundary agents.
- The first implementation plan should start with contract tests around classification, routing, approval ownership, and blocked clinical/privacy outputs.
Limitless OS
v1-agent-contracts
1
Each V1 agent is defined by judgment, memory, tools, queue, output schema, confidence behavior, and human handoff. Implementation must serve this contract rather than reducing agents to CRUD handlers.
- front_office
- content
- command_guardrail
HCS or internal queue item id
07dee66f-8143-4f4d-9af9-b1ed413d2767
- low
- medium
- high
- thought id
- email thread id
- CRM lead id
- schedule ref
- draft ref
short self-contained brief
Seth | Betsy/front office | clinician/performance team | privacy review
memory | decision | task | draft | sop_workflow_rule | question | review
- completed
- drafted
- needs_clarification
- needs_human_approval
- blocked_by_policy
- no_action
what the agent produced or decided
0.0-1.0 plus plain-language reason
memory | decision | task | draft | sop_workflow_rule | question | review
privacy/clinical/automation boundary notes
Seth | Betsy/front office | clinician/performance team | privacy review | none
facts, decisions, examples, or preferences learned
optional queue items created for downstream agents
boolean
agent-specific JSON payload
- Escalate all individualized clinical advice, diagnosis, treatment plan changes, return-to-play implications, and red flags to clinician/performance review.
- Escalate all PHI-sensitive use, patient stories, payer details, staff-sensitive items, and HCS payload concerns to privacy/guardrail review.
- Escalate pricing exceptions, financial disputes, collections language, insurance explanations, and coverage promises to front office/Seth approval.
- Ask a clarifying question when the requested action, owner, patient context, or intended audience is ambiguous.
- Draft first in V1; do not send, publish, schedule, or alter records automatically.
36c57e41-dbda-434b-ba2b-42e13cff1c29
19911d80-32bc-4890-b033-c028aa0c0122
3bbfdbc8-0578-4542-8d64-8f64886b60a2
- thought_capture_agent
- operating_memory_curator
- chief_operating_brain
- intake_concierge
- scheduler
- patient_communication
- content_brain
- social_post_drafter
- human_approval_router
- privacy_clinical_guardrail
- newsletter_agent
- patient_education_agent