Back to skills
SKILL.md
Iblai Vibe Agent Access
CSecurityAdd the agent Access tab (role-based access control for editor and chat roles) to your Next.js app
- 15 stars
- 0 votes
- 0 copies
- 1 view
- Added September 2, 2026
Works with
Security analysis
67/100- Exfiltrates credentials via HTTP — exact pattern from Snyk ToxicSkills study
- Installs packages at runtime which could introduce malicious dependencies
npx -y skills add iblai/vibe --skill iblai-vibe-agent-access --agent claude-codeAre you the author of Iblai Vibe Agent Access?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/iblai-iblai-vibe-agent-access)---
name: iblai-vibe-agent-access
description: Add the agent Access tab (role-based access control for editor and chat roles) to your Next.js app
globs:
alwaysApply: false
metadata:
kind: ui
---
# /iblai-vibe-agent-access
Add the agent **Access tab** -- role-based access management for agents.
Lets admins assign **editor** and **chat** roles, add/remove users and
groups per role, and view existing access policies in a table. This is one
tab in the wider agent-settings family (`access`, `api`, `datasets`,
`disclaimers`, `embed`, `history`, `llm`, `memory`, `prompts`, `safety`,
`settings`, `tools`). All tabs share the same `AgentSettingsProvider`
wrapper.

> **Common setup (brand, conventions, env files, verification):** see [docs/skill-setup.md](https://raw.githubusercontent.com/iblai/vibe/refs/heads/main/docs/skill-setup.md).
## Prerequisites
- Auth must be set up first (`/iblai-vibe-auth`)
- MCP server + skills configured (`@iblai/mcp` in `.mcp.json`)
- `AgentSettingsProvider` must wrap the route (see `/iblai-vibe-agent-setting`
Step 2 if not already set up)
- Ask the user for a real `mentorId` (agent UUID). Do NOT invent one.
## Step 1: Check Environment
Before proceeding, check for an `iblai.env` in the project root. Look for
`PLATFORM`, `DOMAIN`, and `TOKEN` variables. If the file does not exist or
is missing these variables, tell the user:
"You need an `iblai.env` with your platform configuration. Download the
template and fill in your values:
`curl -o iblai.env https://raw.githubusercontent.com/iblai/vibe/refs/heads/main/iblai.env`"
## Step 2: Mount `AgentAccessTab`
The `AgentSettingsProvider` layout from `/iblai-vibe-agent-setting` Step 2 must
already wrap this route. The provider supplies `tenantKey`, `mentorId`,
`username`, and `enableRBAC` via context.
```tsx
// app/(app)/agents/[mentorId]/access/page.tsx
"use client";
import { AgentAccessTab } from "@iblai/iblai-js/web-containers/next";
export default function AgentAccessPage() {
return (
<div className="flex h-full flex-col bg-white">
<AgentAccessTab />
</div>
);
}
```
## Step 3: Customize Labels (Optional)
`AgentAccessTab` renders with the default agent-facing copy
(`AGENT_ACCESS_TAB_LABELS`). Override any string via the `labels` prop:
```tsx
import {
AgentAccessTab,
AGENT_ACCESS_TAB_LABELS,
} from "@iblai/iblai-js/web-containers/next";
<AgentAccessTab
labels={{
header: { title: "Mentor access", description: "Manage who can edit or chat with this mentor." },
table: { roleColumn: "Permission level" },
}}
/>;
```
## Step 4: Use MCP Tools for Customization
```
get_component_info("AgentAccessTab")
get_component_info("AgentSettingsProvider")
```
## `<AgentAccessTab>` Props
Import from `@iblai/iblai-js/web-containers/next`.
| Prop | Type | Required | Description |
|------|------|----------|-------------|
| `labels` | `DeepPartial<AccessTabLabels>` | No | Override user-visible strings (header, table columns, empty states, dialogs) |
| `onPermissionsLoaded` | `(permissions: object) => void` | No | Fired when RBAC permissions for `/users/` and `/groups/` are fetched on mount |
## Related Exports
From `@iblai/iblai-js/web-containers/next`:
- `AGENT_ACCESS_TAB_LABELS` -- the default agent-facing label bundle.
- `AccessTabLabels` -- type for the full label bundle (header, table,
empty states, edit dialog, create dialog, role descriptions).
## Step 5: Verify
Run `/iblai-vibe-ops-test` before telling the user the work is ready:
1. `pnpm build` -- must pass with zero errors
2. `pnpm test` -- vitest must pass
3. Start dev server and touch test:
```bash
pnpm dev &
npx playwright screenshot http://localhost:3000/agents/<id>/access /tmp/agent-access.png
```
## Important Notes
- **Redux store**: Must include `mentorReducer` and `mentorMiddleware`
- **`initializeDataLayer()`**: 5 args (v1.2+)
- **`@reduxjs/toolkit`**: Deduplicated via webpack aliases in `next.config.ts`
- **Peer deps**: `sonner` and `@iblai/iblai-web-mentor` must be installed
(`pnpm add sonner @iblai/iblai-web-mentor`)
- **Shared provider**: `AgentSettingsProvider` must wrap the route at a
layout level. See `/iblai-vibe-agent-setting` Step 2 for the full snippet.
- **RBAC**: The tab checks `enableRBAC` from the provider and
`rbacPermissions` to gate add/edit actions. If `enableRBAC` is `false`
(the default), all actions are permitted.
- **Default roles**: `editor` and `chat`. These come from
`DEFAULT_MENTOR_ROLES` in the package and are not configurable via props.
- **Brand guidelines**: [BRAND.md](https://raw.githubusercontent.com/iblai/vibe/refs/heads/main/BRAND.md)
## RBAC REST API
For custom UI beyond `<AgentAccessTab>`. All endpoints are prefixed with
`${dmUrl}/api/core/` where `dmUrl` is `NEXT_PUBLIC_API_BASE_URL`. RBAC state
is scoped per platform — pass `platform_key={key}` on every request.
### Concepts
- **Resource paths** are hierarchical, rooted at a platform: a policy on
`/platforms/1/mentors/5/` also applies to `/platforms/1/mentors/5/documents/`.
- **Actions** look like `Ibl.Mentor/Mentors/read`, `Ibl.Mentor/Chat/action`,
`Ibl.*` (wildcard).
- **Data actions** gate field-level access:
`Ibl.Mentor/Settings/display_name/read`. Unreadable fields return empty
values; unwritable fields return 403.
- **Policies** bind a Role to Resources for Users and/or Groups. Permissions
are additive across all matching policies. Owner well-known roles
(`mentor-owner`, `document-owner`, etc.) are applied automatically when
the requesting user owns the resource.
### Roles CRUD — `Ibl.Core/Roles/...`
| Method | Path | Action |
|---|---|---|
| GET | `rbac/roles/?platform_key={key}` | `list` |
| POST | `rbac/roles/` | `action` |
| GET | `rbac/roles/{id}/?platform_key={key}` | `read` |
| PUT/PATCH | `rbac/roles/{id}/` | `write` |
| DELETE | `rbac/roles/{id}/?platform_key={key}` | `delete` |
List params: `platform_key` (required), `include_global_roles` (default `true`),
`name`. Body: `{ name, platform_key, actions[], data_actions[] }`.
### Policies CRUD — `Ibl.Core/Policies/...`
| Method | Path | Action |
|---|---|---|
| GET | `rbac/policies/?platform_key={key}` | `list` |
| POST | `rbac/policies/` | `action` |
| GET | `rbac/policies/{id}/?platform_key={key}` | `read` |
| PUT/PATCH | `rbac/policies/{id}/` | `write` |
| DELETE | `rbac/policies/{id}/?platform_key={key}` | `delete` |
List params: `platform_key`, `role_id`, `name`, `username`, `email`, `group`,
`include_users`, `include_groups`. Body: `{ platform_key, name, role_id,
resources[], user_ids[], group_ids[] }`. Resources must start with
`/platforms/` and end with `/`.
### Groups CRUD — `Ibl.Core/Groups/...`
| Method | Path | Action |
|---|---|---|
| GET | `rbac/groups/?platform_key={key}` | `list` |
| POST | `rbac/groups/` | `action` |
| GET | `rbac/groups/{id}/?platform_key={key}` | `read` |
| PUT/PATCH | `rbac/groups/{id}/` | `write` |
| DELETE | `rbac/groups/{id}/?platform_key={key}` | `delete` |
List params: `platform_key`, `owner`, `name`, `username`, `email`,
`include_users`. Body: `{ platform_key, name, unique_id, description, user_ids[] }`.
### Permission check (any authenticated user)
```
POST rbac/permissions/check/
{ "platform_key": "...", "resources": ["/mentors/", "/mentors/42/", "/usergroups/"] }
```
Returns a map of resource → `{ list/create/read/write/delete/chat/... }`.
Accounts for owner roles automatically. Use this to gate UI actions.
### Agent access — `Ibl.Mentor/ShareMentor/...`
| Method | Path | Action |
|---|---|---|
| POST | `rbac/mentor-access/` | `action` |
| GET | `rbac/mentor-access/?platform_key={key}&mentor_id={id}` | `read` |
Body: `{ platform_key, mentor_id, role, users_to_add[], users_to_remove[],
groups_to_add[], groups_to_remove[] }`. Roles: `chat | viewer | editor`.
Substitute `usernames_to_add` / `emails_to_add` if you don't have user IDs.
### Team (UserGroup) sharing — `Ibl.Core/ShareUserGroups/...`
| Method | Path | Action |
|---|---|---|
| POST | `rbac/teams/access/` | `action` |
| GET | `rbac/teams/access/?platform_key={key}&usergroup_id={id}` | `read` |
Body: `{ platform_key, usergroup_id, role, users_to_add[], ... }`. Roles:
`read | edit | view analytics | send notifications`.
### User policy bulk update — `Ibl.Core/UserPolicies/write`
```
PUT platform/users/policies/
[{ "user_id": 101, "platform_key": "...",
"policies_to_add": [...], "policies_to_remove": [...], "policies_to_set": [...] }]
```
Only platform-assignable policies are allowed. `policies_to_set` replaces;
removals run before additions.
### Platform admin toggles (no extra RBAC action)
| Path | Body |
|---|---|
| `POST rbac/student-mentor-creation/set/` | `{ platform_key, allow_students_to_create_mentors }` |
| `GET rbac/student-mentor-creation/status/?platform_key={key}` | – |
| `POST rbac/student-llm-access/set/` | `{ platform_key, llm_resources: ["llms/openai/models/gpt-4o", ...] }` |
| `GET rbac/student-llm-access/status/?platform_key={key}` | – |
### Response permission metadata
Resource responses include a `permissions` object the UI can use to gate
controls:
```json
"permissions": {
"field": { "display_name": { "read": true, "write": true } },
"object": { "delete": false, "write": true }
}
```
Unreadable fields are masked (`""`, `[]`, or `{}`) and `read: false`.
### Resource operations
| Resource | Collection ops | Instance ops |
|---|---|---|
| `mentors` | list, create, chat, web_search, attach_document, voice_record, voice_call, export_chat_history, view_chat_history, view_analytics, view_prompts, share, sell_mentor | read, write, delete + above + show_settings, share_mentor, read_shared_mentor, can_use_embed, view_moderation_logs, view_safety_logs, view_disclaimers |
| `prompts`, `documents`, `tools`, `settings`, `llms`, `mcpservers` | list, create | read, write, delete |
| `usergroups` | list, create | read, write, delete, share_usergroup, read_shared_usergroup |
| `users` | list, write | – |
| `groups`, `policies`, `roles` | list, create | read, write, delete |
| `platforms` | – | can_send_notifications, can_view_analytics, can_manage_users, can_invite |
### Built-in roles
Role names are the platform's seeded display strings, quoted verbatim:
`Tenant Admin` · `Students` · `Mentor Viewer` · `Mentor Editor` · `Mentor Chat` ·
`Student Mentor Creators` · `Analytics Viewer` · `Notification Manager` ·
`Enrollment Manager` · `LLM Users` · `LLM Model Access` · `List Users` · `List Teams` ·
`Create Teams` · `Read Team` · `Edit Team` · `Billing Manager`.
Attribution
Comments
Loading comments…
Newsletter
Get the weekly list of safe new skills, and the ones we pulled.