BusinessRolesAI
AI Agent Roles
This file indexes AI Agent roles without duplicating the full agent scope
Purpose
This file indexes AI Agent roles without duplicating the full agent scope
documents under docs/development/ai-control-plane/agent-scopes.
AI Agent roles are scoped helpers. They do not own final decisions, access, legal commitments, merge decisions, or production deployment approvals.
Existing Scope Docs
- Architect Agent
- Docs Agent
- UI Agent
- Shell Agent
- App Agent
- Registry Agent
- API Agent
- QA Agent
- Integrator Agent
- Business Agent
- Growth Agent
- Content Agent
- Ops Agent
Join Us Agent Map
The Join Us app can display these AI agent hats beside human roles. They are role-scoped collaborators, not autonomous workers with production authority.
| Agent | Mission | Boundary | Human reviewer |
|---|---|---|---|
| Docs Agent | Keep docs structure, links, summaries, and AI-readable context healthy. | Does not own design tokens, launch copy, or legal conclusions. | AI Workflow Librarian |
| Design System Agent | Review tokens, visual rules, component anatomy, and design doctrine. | Does not own launch campaigns or runtime data. | Founding Frontend Architect |
| Storybook Agent | Prepare stories, state matrices, and visual regression references. | Does not own Product Hunt assets or pricing strategy. | Founding Frontend Architect |
| Deck Agent | Turn approved thesis into deck outlines and deck-as-code drafts. | Does not invent traction, financials, or legal claims. | Founder / Product Architect |
| Launch Agent | Map launch checklists, release assets, waitlist steps, and external channels. | Does not own long-term community moderation. | Growth & Community Lead |
| Growth Agent | Draft experiments, content angles, channel hypotheses, and metrics. | Does not make paid acquisition or legal claims. | Growth & Community Lead |
| Governance Agent | Normalize roadmap, RFC, ADR, status, and role objects for visibility. | Does not decide merge, deploy, or legal status. | Founder / Product Architect |
| Compliance Agent | Maintain compliance checklists and operational maps. | Does not provide legal, tax, immigration, or regulatory advice. | Ops Steward |
| Data Agent | Map personal, app, source, and context data models at planning level. | Does not implement migrations or store sensitive data. | Founder / Product Architect |
| App Registry Agent | Review app manifest, registry, launchability, and route consistency. | Does not approve marketplace apps or billing rules. | Registry Agent / Founder |
| Community Spaces Agent | Map community Space roles, publishing ideas, and contributor flows. | Does not run community moderation or public policy alone. | Growth & Community Lead |
| Developer Platform Agent | Draft SDK, CLI, examples, and developer docs direction. | Does not create packages before app contract scope is approved. | Founding Frontend Architect |
Frontend Package Agent Hats
One human Frontend Package Steward can review the frontend package area while
AI agent hats stay narrower for recurring package maintenance, dependency
tracking, Storybook coverage, and policy checks.
| Agent | Mission | Boundary | Human reviewer |
|---|---|---|---|
| Data Contract Agent | Keep @prox-os/data-contract types, examples, and docs aligned. | Does not add UI or storage behavior to the contract package. | Frontend Package Steward |
| Data Table Agent | Track table/grid primitives, TanStack Table usage, and Glide previews. | Does not build a full Airtable or spreadsheet clone. | Frontend Package Steward |
| Visualization Agent | Track ECharts, Vega-Lite, Recharts, chart stories, and chart policies. | Does not make one chart engine the only supported direction. | Frontend Package Steward |
| Actions Package Agent | Track shared action intent, common actions, action UI primitives, and command/context-menu integration docs. | Does not execute Shell actions, backend actions, or permission decisions. | Frontend Package Steward |
| Command Package Agent | Track command contracts, command palette primitives, provider docs, ranking helpers, and action-command bridges. | Does not rewrite the OS Shell launcher or bind global shortcuts. | Frontend Package Steward |
| Forms Package Agent | Track form contracts, validation adapters, field primitives, schema forms, Storybook examples, and backend form contract docs. | Does not build a low-code form builder or wire real backend submission. | Frontend Package Steward |
| Editor Package Agent | Track editor contracts, Tiptap/CodeMirror adapters, editor primitives, content safety docs, and AI/collaboration integration notes. | Does not build a Notion clone, wire real collaboration, or call AI/backend services. | Frontend Package Steward |
| Collaboration Package Agent | Track collaboration contracts, presence, comments, threads, mentions, share panels, review approval, provider boundaries, and stories. | Does not wire realtime backends, notifications, search, activity, security policy, or Shell routing. | Frontend Package Steward |
| Notifications Package Agent | Track notification contracts, badges, bells, toasts, notification center, digests, preferences, adapter examples, and stories. | Does not wire push services, service workers, delivery backends, activity timelines, or Shell routing. | Frontend Package Steward |
| Search Package Agent | Track search contracts, local adapter, filters, facets, highlights, previews, saved searches, index stats, adapter examples, and stories. | Does not wire backend search providers, crawlers, embeddings, vector databases, command palette runtime, or Shell routing. | Frontend Package Steward |
| Files Package Agent | Track file contracts, storage provider refs, browser/picker primitives, previews, transfer states, adapter examples, and stories. | Does not wire real storage SDKs, uploads, downloads, browser file-system permissions, sync engines, or Shell routing. | Frontend Package Steward |
| AI UI Package Agent | Track AI session/message/run/tool-call/approval/artifact/context/citation contracts, mock data, adapter examples, and stories. | Does not wire LLM providers, AI SDKs, MCP, streaming backends, agent runtimes, tool execution, RAG, or Shell routing. | Frontend Package Steward |
| Security UI Package Agent | Track security principal/resource/permission/risk/policy/audit/session/token/access review contracts, mock data, adapter examples, and stories. | Does not wire auth providers, policy engines, KMS, scanners, real audit persistence, or permission enforcement. | Frontend Package Steward |
| Workspace UI Package Agent | Track workspace/space/member/role/resource/project/settings/onboarding/integration contracts, mock data, adapter examples, and stories. | Does not wire workspace backends, member invite delivery, billing, project management engines, permission enforcement, or Shell routing. | Frontend Package Steward |
| Activity Package Agent | Track activity actor/target/object/event/feed/group/digest/stats contracts, mock data, timeline primitives, adapter examples, and stories. | Does not wire analytics SDKs, telemetry, realtime streams, event stores, audit persistence, or Shell routing. | Frontend Package Steward |
| Media Package Agent | Track media resources, embeds, transforms, allowlists, and stories. | Does not own uploads, storage, or OS Shell context menus. | Frontend Package Steward |
| Maps Package Agent | Track MapLibre, map contracts, tile policy, attribution, and stories. | Does not run geocoding, routing, or production tile services. | Frontend Package Steward |
| Frontend Package Docs Agent | Keep docs/platform/packages/frontend concise, current, and cross-linked. | Does not invent package commitments beyond implemented scope. | Frontend Package Steward |
| Storybook Coverage Agent | Fill package story gaps for default, loading, empty, and error states. | Does not replace package typecheck, lint, or runtime QA. | Frontend Package Steward |
| License & Policy Watch Agent | Watch third-party licenses, tile usage policy, attribution, and risk notes. | Does not provide legal advice or approve production compliance. | Frontend Package Steward |
Backend Package Agent Hats
The human role is Backend Platform Steward. Backend AI hats can be more
granular so future CodeX/Cursor workers can generate contracts, fixtures, tests,
and docs without owning production authority.
| Agent | Mission | Boundary | Human reviewer |
|---|---|---|---|
| API Contract Agent | Maintain OpenAPI, route schemas, response envelopes, and error models. | Does not implement runtime middleware or DB access. | Backend Platform Steward |
| API Kit Agent | Review Hono helper boundaries, request context, middleware order, and error handling. | Does not own business routes. | Backend Platform Steward |
| DB Schema Agent | Guard schema conventions, migrations, seed data, and local fixtures. | Does not run production migrations. | Backend Platform Steward |
| Auth Flow Agent | Map identity, sessions, OAuth, anonymous upgrade, and token rotation. | Does not own permission policy. | Backend Platform Steward |
| Permission Policy Agent | Review scopes, grants, resource policies, app permissions, and agent tool approvals. | Does not own auth identity or billing plans. | Backend Platform Steward |
| Storage Adapter Agent | Track R2/object storage contracts, signed URLs, object keys, and lifecycle notes. | Does not own image processing or permission policy. | Backend Platform Steward |
| Media Service Agent | Track variants, transforms, EXIF cleanup, thumbnails, metadata, and provenance. | Does not own frontend media UI. | Backend Platform Steward |
| Search Index Agent | Map search contracts, indexing jobs, facets, ranking, and permission-aware results. | Does not bypass permissions or build frontend search UI. | Backend Platform Steward |
| Jobs Workflow Agent | Review queues, retries, idempotency, cron, dead letters, and recovery. | Does not own domain policy. | Backend Platform Steward |
| Webhook Integration Agent | Guard signatures, provider fixtures, event replay, idempotency, and routing. | Does not perform direct business side effects. | Backend Platform Steward |
| Agent Runtime Agent | Map agent runs, tool calls, MCP bridge, artifacts, permissions, and human gates. | Does not execute tools without permission and audit. | Backend Platform Steward |
| Audit & Compliance Agent | Review audit vocabulary, actor/resource refs, retention, and export readiness. | Does not provide legal advice or replace observability. | Backend Platform Steward |
| Billing Agent | Map plans, tips, subscriptions, entitlements, usage limits, and payment replay. | Does not connect live payment providers alone. | Backend Platform Steward |
| Observability Agent | Track logs, traces, metrics, request IDs, agent telemetry, and dashboards. | Does not replace audit or send user notifications. | Backend Platform Steward |
| Test Harness Agent | Ensure backend PRs include fixtures, mocks, contract tests, and rollback notes. | Does not approve missing test coverage. | Backend Platform Steward |
Additional Governance Agent Hats
| Agent | Mission | Human reviewer | Prompt/run location |
|---|---|---|---|
| RFC Triage Agent | Decide whether a raw idea needs RFC, ADR, task graph, or simple patch. | Founder / Product Architect | docs/development/ai-control-plane/prompt-templates / docs/development/ai-control-plane/runs |
| ADR Scribe Agent | Draft concise decision records from approved decisions. | Founder / Product Architect | docs/development/adr, docs/platform/architecture/decisions |
| Prompt Librarian Agent | Compress raw prompts into reusable summaries and templates. | AI Workflow Librarian | docs/development/ai-control-plane/prompt-templates |
| Code Review Assistant Agent | Read diffs and surface bugs, boundary violations, and missing checks. | Integrator Agent or owning human role | Integration reports |
| Performance Scout Agent | Inspect bundle, render, and latency risks. | Founding Frontend Architect | docs/operations/performance |
| Security Scout Agent | Review access, secrets, AI work product, and contributor permission risks. | Ops Steward | docs/trust/security |
| China Readiness Research Agent | Gather China readiness research notes without legal conclusions. | China Readiness Lead | docs/trust/readiness/apac/china-readiness.md |
| Growth Memo Agent | Draft market, community, launch, and content memos. | Growth & Community Lead | docs/go-to-market/how-prox-os-grows |
| Deck Outline Agent | Turn approved thesis into deck outlines. | Founder / Product Architect | docs/go-to-market/decks |
| Roadmap Curator Agent | Normalize roadmap status and links across docs and UI. | Founder / Product Architect | docs/product/roadmap |
| Governance Console Assistant Agent | Prepare governance object summaries for the OS UI mock. | Ops Steward | docs/business/governance |
Forbidden Actions
- No secrets handling.
- No production deploy.
- No payment, legal, tax, immigration, or hiring commitment.
- No real account access without explicit human approval outside repo docs.
- No direct API integration unless a task graph explicitly permits it.