Resources, skills, and prompts

MCP resources explain RevoEngine semantics that do not fit in a tool schema. Prompts provide reusable starting messages for common operating modes. A capable client should read them before guessing how a platform concept works.

Recommended orientation sequence

Read effective context

Start with revo://context/effective to understand the catalogue visible in the authenticated session.

List skills

Read revo://skills, select the skill matching the task, then read revo://skills/{skillId}.

Load exact declarations when coding

For low-code work, inspect revo://runtime/low-code-declarations and read the referenced type declaration instead of inventing an API surface.

Discover tools last

Use tools/list after the client understands the resource model, safety expectations, and relevant runtime contract.

Core resources

URI	Content	Use it when
`revo://platform/about`	Platform purpose and authority order	The client needs a concise product orientation
`revo://platform/resource-model`	Control-plane resources and execution relationships	Planning work across Components, Endpoints, Databases, Storage, or Automation
`revo://platform/runtime-surfaces`	Runtime globals, execution surfaces, and library access	Writing or reviewing low-code runtime logic
`revo://platform/enterprise-standards`	Governance, evidence, approvals, and safe-change expectations	Preparing a production mutation or operational response
`revo://context/effective`	Session-visible catalogue summary	Verifying what the authenticated client can currently see
`revo://skills`	Searchable skill catalogue with resource links	Selecting task-specific RevoEngine guidance
`revo://runtime/low-code-declarations`	Index of current declaration resources	Locating exact editor-compatible TypeScript contracts
`revo://types/api.public.d.ts`	Public low-code declaration text	Generating code against `api`, `storage`, `agent`, `util`, `inject(...)`, and related globals

The catalogue may add resources over time. Use resources/list and resources/templates/list rather than treating this table as a protocol negotiation substitute.

Read a resource

{
  "jsonrpc": "2.0",
  "id": "resource-1",
  "method": "resources/read",
  "params": {
    "uri": "revo://platform/resource-model"
  }
}

The result contains a contents array with the resource URI, MIME type, and text. JSON resources are returned as text with application/json; parse them only after a successful response.

Resource templates

RevoEngine currently exposes templates for:

URI template	Resolves to
`revo://skills/{skillId}`	Full guidance for one skill selected from `revo://skills`
`revo://types/{name}`	A named public runtime declaration resource

Do not guess a skill id or declaration filename. Resolve it from the corresponding index resource first.

Skills are guidance, not execution

A RevoEngine skill describes how to approach a product workflow, what evidence to collect, which capability families are relevant, and which safety boundaries matter. Reading a skill does not run a tool and does not grant permissions. This separation lets a client load detailed domain guidance without expanding its execution authority. Only skills explicitly classified for the public MCP catalogue are listed. Internal Harness guidance, private prompts, source-reference pointers, capability-routing policy, and model-only diagnostic material are not published through this resource. A public skill explains the supported product contract and safe usage; it is not a copy of RevoEngine’s internal execution design.

Select a workflow before loading its instructions

The version 2 skill catalogue provides complete selection descriptions, resource URIs, relevant capability names, and optional requires identifiers for prerequisite guidance. Select the narrow workflow first: querying Database rows, changing a Database definition, importing data, exporting a file, investigating one failure, and reporting execution statistics are separate tasks. Resource and investigation responses may include relatedSkills. These are optional recommendations, not loaded instructions, additional permissions, or a requirement to read every related workflow. Resolve selected identifiers through the catalogue and read their resources together with any missing prerequisites. Reuse complete guidance already present in the client context; routine metadata reads do not require loading an entire dependency tree. Discovery returns canonical identifiers. Retired skill identifiers remain readable compatibility aliases, so existing saved resource links can still resolve. Direct MCP calls use the authenticated client’s platform authority. They do not pause for Interactive Agent approval; clients must not assume that an approval step described in an Agent workflow protects an MCP mutation.

Reusable prompts

Use prompts/list to discover the current prompt catalogue. The present workflow families include:

Prompt	Purpose
`revo.platform_orientation`	Explain RevoEngine and route an unfamiliar request to the correct product surface
`revo.enterprise_change_safety`	Prepare a mutation with blast-radius, approval, evidence, and rollback considerations
`revo.control_plane_workflow`	Discover, inspect, change, and verify control-plane resources
`revo.low_code_runtime_guidance`	Work with runtime globals, Components, libraries, Storage, and Sandbox validation

Retrieve one prompt with:

{
  "jsonrpc": "2.0",
  "id": "prompt-1",
  "method": "prompts/get",
  "params": {
    "name": "revo.enterprise_change_safety"
  }
}

The result is a standard MCP prompt message collection. The client decides how to present or apply it.

Independent client context

The public MCP initialization instructions are server-level operating guidance. They do not copy a user’s Assistant preferences, personal memory, private Agent workspace, or active Assistant thread into an independent MCP client. If a workflow needs Assistant conversation state, use the Assistant surface. If it needs direct instance operations from another client, use public MCP with an identity and context appropriate to that client.

Integrate

Model Context Protocol

Node.js SDK

CLI

VS Code

Resources, skills, and prompts

Recommended orientation sequence

Core resources

Read a resource

Resource templates

Skills are guidance, not execution

Select a workflow before loading its instructions

Reusable prompts

Independent client context

Assistant and Agent memory

Low-code reference

​Recommended orientation sequence

​Core resources

​Read a resource

​Resource templates

​Skills are guidance, not execution

​Select a workflow before loading its instructions

​Reusable prompts

​Independent client context

Assistant and Agent memory

Low-code reference

Recommended orientation sequence

Core resources

Read a resource

Resource templates

Skills are guidance, not execution

Select a workflow before loading its instructions

Reusable prompts

Independent client context