Model skills library
Installable business playbooks you can put into a tenant with Trace AI. Each item is a playbook (resource maps, stages, quality gates, definition of done), not a prompt. Canonical records live in DataPool schemas. Default notifications are in-app.
Two families:
- B2B value chain — Marketing → Sales → Delivery → Support.
- AI adoption advisor — discovery, scoring, and a vendor-neutral roadmap; optional later bridge onto Tealfabric drafts. End-user flow: AI adoption advisor.
Document information
| Field | Value |
|---|---|
| Canonical URL | /docs/07_ai-agents-and-mcp/model_skills/00_index |
| Version (published date) | 2026-08-27 |
| Tags | ai, skills, playbooks, datapool, trace-ai |
How to install
In Trace AI, ask to install a named playbook from this library (for example “Install tenant-data-model from the model skills library”). Trace loads load_platform_skill with skill_id: model-skill-import, fetches this catalog, resolves Prerequisites against the tenant (DataPool schemas, recipient users), then writes SKILLS/{skill_id}/.
Trace searches the tenant first. If a Prerequisite is missing or ambiguous, it asks you and may propose a default. It does not assume Mattermost or any other chat product exists. Installed playbooks must contain bound values, not leftover {{tokens}}.
Blank-slate authoring (no library page) still uses skill_id: tenant-skill-authoring.
Value chain
Marketing → Sales → Delivery → Customer Support
\ \ \
handovers (own playbooks)
Shared foundation (tenant-data-model, crm-field-mapping) must be installed before stage playbooks that read or write those DataPool schemas.
| Stage | suggested_skill_id | Kind | Prerequisites (summary) | Page |
|---|---|---|---|---|
| Shared | tenant-data-model | Resource / canonical model | Bind or create DataPool schemas | tenant-data-model |
| Shared | crm-field-mapping | Resource guardrails | crm_integration_id; depends on tenant-data-model | crm-field-mapping |
| Marketing | marketing-lead-qualification | Procedure | leads schema | marketing-lead-qualification |
| Sales | sales-process | Procedure | opportunities schema | sales-process |
| Sales | rfp-handling | Procedure | rfp_requests; requirements + product-docs fit before offer; CRM master | rfp-handling |
| Delivery | delivery-process | Procedure | deliveries schema | delivery-process |
| Customer Support | support-process | Procedure | support_cases schema | support-process |
| Handover | marketing-to-sales-handover | Handover | handovers schema; in-app recipient | marketing-to-sales-handover |
| Handover | sales-to-delivery-handover | Handover | handovers schema; in-app recipient | sales-to-delivery-handover |
| Handover | delivery-to-support-handover | Handover | handovers schema; in-app recipient | delivery-to-support-handover |
Paired Autonomic Agent templates: Model agents library.
AI adoption advisor
Assessment playbooks do not create ProcessFlow, WebApps, or Autonomic Agents. ai-tealfabric-bridge is a separate, opt-in step.
| Stage | suggested_skill_id | Kind | Prerequisites (summary) | Page |
|---|---|---|---|---|
| Shared | ai-adoption-data-model | Resource / canonical model | Bind or create DataPool schemas | ai-adoption-data-model |
| Assessment | ai-discovery | Procedure | Strategy via get_platform; as-is schemas | ai-discovery |
| Assessment | ai-opportunity-assessment | Procedure | ai_use_cases schema | ai-opportunity-assessment |
| Assessment | ai-roadmap | Procedure | roadmap_items schema | ai-roadmap |
| Assessment | ai-operating-model | Procedure | Engagement folder | ai-operating-model |
| Vertical | ai-hr-backoffice | Resource / catalog | as_is_processes schema | ai-hr-backoffice |
| Bridge (opt-in) | ai-tealfabric-bridge | Procedure | implementation_links; never run as part of discovery | ai-tealfabric-bridge |
User guide (install, workshop, JWT board pack, bridge): AI adoption advisor.
What this library is not
- Platform MCP playbooks (
load_platform_skillids such asdatapool,entities,notifications) — those stay in the productSKILLS/tree. - Persona or style prompts.
- Skills an agent can invent without org knowledge (generic “search the CRM”).