Gabriel Operator
Gabriel Operator v1.5.1
Publisher description
From the marketplace listing
Tell this plugin what your AI persona should do. It asks a few questions, then builds a complete end-to-end persona on Gabriel Operator: research, tools, prepared actions, your approval, then the real workflow. Ready to chat, try, and improve - not a stack of disconnected agents. Please Sign up at https://gabrieloperator.com/signup. Open Workspace, go to Dashboard, and copy your Gabi token (the key that starts with gabi_). Connect that token in ChatGPT before you run this plugin. Do not paste the raw token into chat.
Language: English · Automatically detected from descriptions.
Files & skills
File archives
Skill instructions
persona-builder7.63 KB
---
name: persona-builder
description: Interview the user and author a local Gabriel AI Persona draft covering page copy, list schema, pipeline stages, optional team-agent graph, and optional review-only workflow. Use when the user wants to design a new persona in files they control. Produces local drafts only and does not publish, provision, authenticate, or call remote services.
---
# Persona Builder
Design one Gabriel AI Persona as local draft files for human review.
This public skill covers the same draft files as `digital-twin-page`,
`list-builder`, `pipeline-builder`, and `team-agents`. It does not create live
Gabriel resources. Live create, git bind, and publish stay in Gabriel Operator
after the user reviews these files.
## Public authoring boundary
- Work only in the repository the user explicitly selects.
- Write local draft files only. Do not commit, push, publish, deploy, or run.
- Do not request, infer, copy, or store passwords, API keys, access tokens, cookies, or session data.
- Do not call remote APIs, open Gabriel Operator in a browser, or sync live records.
- Do not add real personal, confidential, financial, or authentication data.
- Use placeholder ids such as `page_example`. Do not invent production database ids.
- If the user asks to provision, publish, or authenticate, refuse and point them to Gabriel Operator.
## Interview
Collect, propose, confirm, then write files. Do not dump empty JSON and stop.
1. Ask what the persona does if the user did not already describe it.
2. Propose a title, short description, visibility (`private` default), list columns, pipeline stages, and whether a team-agent graph or a review-only workflow is needed.
3. Confirm the proposal.
4. Write the files below.
5. Summarize the files changed and note that live create/publish is a separate user action.
## Canonical files
```text
assets/chat-config.json
assets/list.json
assets/pipeline.json
assets/team-agent.json
assets/task-orchestration.json
assets/workflow.json
```
Write `chat-config.json`, `list.json`, and `pipeline.json` for every new persona.
Write team-agent and workflow files only when the confirmed proposal includes them.
Preserve unknown fields and stable ids when a file already exists.
## Page copy (`assets/chat-config.json`)
Supported fields only: `pageProfile.title`, `pageProfile.description`,
`pageProfile.longDescription`, `pageProfile.profilePicture`,
`pageProfile.bannerImage`, `pageProfile.bannerType`, `pageProfile.tags`,
`pageProfile.category`, `pageProfile.subcategory`, `publishedConfig.name`,
`publishedConfig.firstMessage`, `publishedConfig.systemPrompt`.
Keep conversational instructions clear and non-deceptive. Do not claim tools,
connectors, or live data this draft cannot provide.
```json
{
"schemaVersion": 1,
"pageId": "page_example",
"publishedConfig": {
"name": "Assistant",
"firstMessage": "How can I help?",
"systemPrompt": "Answer using the material supplied by the user."
},
"pageProfile": {
"title": "Assistant",
"description": "A helpful digital twin",
"tags": []
},
"updatedAt": 0
}
```
## List schema (`assets/list.json`)
Edit schema only. Do not import or write live records. Column types: `text`,
`number`, `boolean`, `date`, `select`. Use `options` only on `select` columns.
Additional lists, if requested, go in `assets/lists/<slug>.json` with the same shape.
```json
{
"schemaVersion": 1,
"listId": "list_example",
"list": {
"id": "list_example",
"name": "Example List",
"description": "Review-only list schema",
"collectionId": "collection_example",
"columns": [
{
"key": "title",
"label": "Title",
"type": "text",
"required": true
}
]
},
"commitMessage": "Draft list schema"
}
```
## Pipeline (`assets/pipeline.json`)
One pipeline is one state-machine draft. Stages are metadata. Transitions are
allowed moves. Do not run the machine or modify live records.
```json
{
"schemaVersion": 1,
"pageId": "page_example",
"pipelineId": "pipe_example",
"pipeline": {
"id": "pipe_example",
"pageId": "page_example",
"name": "Example Pipeline",
"collectionId": "collection_example",
"columns": [
{
"id": "col_title",
"key": "title",
"label": "Title",
"type": "text",
"required": true,
"order": 1
}
],
"stages": [
{
"id": "stage_new",
"name": "New",
"type": "start",
"description": "Initial stage"
},
{
"id": "stage_done",
"name": "Done",
"type": "end",
"description": "Completed stage"
}
],
"transitions": [
{
"id": "tr_start_to_done",
"name": "Complete",
"fromStageId": "stage_new",
"toStageId": "stage_done",
"trigger": "manual"
}
]
},
"taskIds": [],
"commitMessage": "Draft pipeline definition"
}
```
## Team agent (optional)
Edit `assets/team-agent.json` and `assets/task-orchestration.json` only.
Use declarative local nodes and edges. Do not add browser automation, desktop
control, shell execution, credentials, or live connector secrets.
```json
{
"appId": "app_example",
"endpointId": "endpoint_example",
"endpoint": {
"id": "endpoint_example",
"slug": "example-endpoint",
"name": "Example Endpoint",
"workspace": {
"nodes": [
{
"id": "node_start",
"type": "start",
"label": "Start",
"config": {
"inputSchema": {
"type": "object",
"properties": {}
}
}
},
{
"id": "node_output",
"type": "output",
"label": "Output",
"config": {}
}
],
"edges": [
{
"id": "edge_start_output",
"source": "node_start",
"target": "node_output"
}
]
}
},
"commitMessage": "Draft team-agent endpoint"
}
```
```json
{
"schemaVersion": 1,
"mappings": [],
"commitMessage": "Draft task orchestration"
}
```
## Workflow (optional)
Edit `assets/workflow.json` only. Supported review-only steps: `navigate`
(with `disableBrowser: true` and empty `url`), `wait`, `llm`, `confirmation`,
`api_output`. If the user needs any other action, describe it in the handoff
instead of adding it.
Keep every step field at the step root. Give each step a unique `stepId` and
sequential `step_number`.
## Finish
Validate JSON syntax, unique ids, and transition/edge references locally.
Report the files changed. Tell the user these are drafts until they create and
publish the persona in Gabriel Operator.
## Live create in Gabriel Operator
This review pack does not collect tokens or call Gateway. For a live persona:
1. Sign up at https://gabrieloperator.com/signup
2. Open **Workspace → Dashboard** and copy a Gabi token from the **Gateway API key** pill
3. Connect GitHub on the same account
4. Connect that token to ChatGPT/Codex MCP (`Authorization: Bearer gabi_…`) or `GABRIEL_TOKEN`
5. Use the live Persona Builder pack, not this local-draft skill
Do not request, store, or use the token inside this skill.
## UI fallback for live configure
If the user needs a live setting this pack cannot draft in files (Composio keys,
Arcade/Nango/Scalekit, GitHub connect, voice BYOK), send them to the UI after a
real page exists. Do not collect those secrets here.
Personal workspace (same destination as **Configure** on the persona page):
https://gabrieloperator.com/workspace/edit-persona/{pageId}
- Composio keys: `?tab=simulated-world` then expand **MCP connectors**
- GitHub: https://gabrieloperator.com/workspace/developer-settings
Give the full URL. Name the tab and the control.
Referenced files: 1
Package details
Publisher declarations from the archived package. These are separate from our research and the live service's terms.
- Package license
- MIT
- Package author
- Gabriel Operator
- Keywords
- gabriel-operator, persona-builder, digital-twin, skills
Declared capabilities
- Read
- Write
Package observed Oct 2, 2026.
Technical details
- First seen
- Sep 30, 2026 · 22:02 UTC
- Last seen
- Oct 2, 2026 · 18:00 UTC
- Collection status
- Collected
plugins_6a8459174a4c819182de883ba421ab02
Download plugin data (JSON)