integrate-fusion-agent
Skill6 starsChanged 10 days agoApache-2.0View on GitHub
MUST be used when adding AI, Atlas, an agent, a chat UI, or LLM features to a Flows/Fusion app. Use the Atlas/EOS sidebar via @cognite/app-sdk — not useAtlasChat, vendored atlas-agent, or per-row chat completions. Triggers: atlas, EOS, PAIA, agent chat, chat UI, sendAgentMessage, sendAgentLayoutMode, registerAgentServer, connectToHostApp, useAtlasChat, LLM. In-app chat: integrate-atlas-chat only if the host sidebar cannot work.
- Installs packages
In this repository34 other files it gives its agents.
- aura-reviewSkill
- call-cognite-functionSkill
- code-qualitySkill
- connect-atlas-mcpSkill
- correctness-and-error-handlingSkill
- create-agentSkill
- create-client-toolSkill
- dependencies-auditSkill
- designSkill
- dm-graph-traversalSkill
- dm-limits-and-best-practicesSkill
- flows-app-briefSkill
- flows-code-reviewSkill
- flows-design-reviewSkill
- flows-external-app-submitSkill
- flows-review-checksSkill
- graph-viewerSkill
- handle-startup-stateSkill
- hide-show-shellSkill
- integrate-atlas-chatSkill
- integrate-file-viewerSkill
- integrate-todo-listSkill
- migrate-app-to-flowsSkill
- performanceSkill
- pull-changes-resolve-conflictsSkill
- reveal-3dSkill
- securitySkill
- setup-flows-authSkill
- setup-python-toolsSkill
- skill-creatorSkill
- start-flows-appSkill
- test-coverageSkill
- tygenSkill
- use-topbarSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 15
Default AI path: the platform Atlas sidebar (EOS / Fusion PAIA) via @cognite/app-sdk.
- Distinctive: line 32
Ask which of the three capabilities are needed.
- Very specific: line 66
Call at the root; pass api down or via context.
- Very specific: line 238
Do not map chat completions (Atlas agents/chat, OpenAI/Anthropic, useAtlasChat().send) over DMS/SDK rows.
- Distinctive: line 240
If per-item completions are an explicit product requirement (default: no):
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Integrate Atlas / EOS Sidebar
Default AI path: the platform Atlas sidebar (EOS / Fusion PAIA) via @cognite/app-sdk. Do not embed useAtlasChat, vendor atlas-agent, or call third-party LLM APIs.
integrate-atlas-chat only if the user explicitly requires in-app chat and connectToHostApp cannot provide the sidebar (standalone app; always rejects). There is no manifest field for this.
Implement only what is needed:
- Open — Topbar Atlas button;
sendAgentLayoutModefor in-app triggers - Message —
sendAgentMessageto inject context - Server — resources (app state) and actions (tools)
Step 0 — Read the app
package.json— package manager,@cognite/app-sdksrc/App.tsx— structure, existing SDK usage
Ask which of the three capabilities are needed. Do not offer an in-app chat unless they already insisted.
Step 1 — Install
pnpm add @cognite/app-sdk (or npm/yarn). Minimum 0.3.1.
Step 2 — Connect to the host
connectToHostApp rejects outside Fusion (standalone vite dev). Catch that; hide agent triggers when api is null.
Comlink proxies are callable — setApi(proxy) makes React treat the proxy as an updater and stores a Promise. Always setApi(() => resolvedApi).
// src/hooks/useHostApp.ts
import { useState, useEffect } from 'react';
import { connectToHostApp, type HostAppAPI } from '@cognite/app-sdk';
export function useHostApp(): HostAppAPI | null {
const [api, setApi] = useState<HostAppAPI | null>(null);
useEffect(() => {
connectToHostApp({ applicationName: 'my-app' })
.then(({ api: resolvedApi }) => setApi(() => resolvedApi))
.catch(() => { /* outside Fusion — no-op */ });
}, []);
return api;
}Call at the root; pass api down or via context. typeof proxy.method === 'function' is always true — do not feature-detect with typeof; use try/catch.
Step 3 — Open the sidebar
Primary launcher: Aura Topbar Atlas (systemActions.atlas.visible: true, see use-topbar). No second "Open Assistant" control.
sendAgentLayoutMode is for contextual triggers only (sidebar | fullscreen | closed):
await api.sendAgentLayoutMode({ mode: 'sidebar' });Step 4 — Send a message
Pair with sendAgentLayoutMode. newSession: true for a new task from an item; omit to continue the thread. Put names/IDs/state in the message — one sidebar turn, not N completions over query rows.
await api.sendAgentLayoutMode({ mode: 'sidebar' });
await api.sendAgentMessage({
message: `Analyse the schedule for "${itemName}" and suggest how to reduce total duration.`,
newSession: true,
});Step 5 — Agent server
Register on mount, unregister on unmount. Factories take services as args so they can be unit-tested without React:
src/features/agent/
agentActions.ts — (deps) => Action[]
agentResources.ts — (deps) => Resource[]
useAgentServer.ts — register / unregisterResource read() returns { type: 'json', data } (preferred) or { type: 'text', text }. Write description like a docstring.
// src/features/agent/agentResources.ts
import { createAgentResource } from '@cognite/app-sdk';
export function buildAgentResources(storage: StorageService) {
return [
createAgentResource({
uri: 'my-app://current-state',
name: 'Current application state',
description:
'Items currently visible, their statuses, and active filters. Read before answering questions about what the user is looking at.',
async read() {
return [{ type: 'json', data: storage.getAll() }];
},
}),
];
}Actions: snake_case names, Zod params, .describe() on every field. The agent does not confirm before calling — mutating actions must say so in description and require prior user approval.
// src/features/agent/agentActions.ts
import { createAgentAction } from '@cognite/app-sdk';
import { z } from 'zod';
export function buildAgentActions(dataService: DataService) {
return [
createAgentAction({
name: 'get_item_details',
description: 'Full details for an item by ID, including history.',
parameters: z.object({
item_id: z.string().describe('The ID of the item to retrieve'),
}),
async handler({ item_id }) {
const item = await dataService.getItem(item_id);
return { content: [{ type: 'json', data: item }] };
},
}),
];
}createAgentAction({
name: 'update_item_status',
description:
'Update item status. Call ONLY when the user has explicitly approved the change.',
parameters: z.object({
item_id: z.string().describe('The item to update'),
status: z.enum(['active', 'closed', 'pending']).describe('The new status'),
}),
async handler({ item_id, status }) {
storage.updateStatus(item_id, status);
return { content: [{ type: 'json', data: { success: true } }] };
},
})// src/features/agent/useAgentServer.ts
import { useEffect } from 'react';
import { createAgentServer, registerAgentServer, type HostAppAPI } from '@cognite/app-sdk';
import { buildAgentActions } from './agentActions';
import { buildAgentResources } from './agentResources';
import { useStorageService } from '../storage/StorageServiceContext';
import { useDataService } from '../data/DataServiceContext';
export function useAgentServer(api: HostAppAPI | null): void {
const storage = useStorageService();
const dataService = useDataService();
useEffect(() => {
if (!api) return;
const server = createAgentServer({
uri: 'my-app', // Fusion namespaces with instance ID
actions: buildAgentActions(dataService),
resources: buildAgentResources(storage),
});
void registerAgentServer(api, server).catch((err: unknown) => {
console.warn('[agent] registerAgentServer failed:', err);
});
return () => {
void api.unregisterAgentServer('my-app').catch((err: unknown) => {
console.warn('[agent] unregisterAgentServer failed:', err);
});
};
}, [api, storage, dataService]);
}Step 6 — Wire together
function App() {
const api = useHostApp();
useAgentServer(api);
return (
<AppLayout>
<MainContent onAnalyseItem={async (item) => {
if (!api) return;
await api.sendAgentLayoutMode({ mode: 'sidebar' });
await api.sendAgentMessage({
message: `Analyse "${item.name}" (id: ${item.id}).`,
newSession: true,
});
}} />
</AppLayout>
);
}Test factories directly:
const [getItemAction] = buildAgentActions({
getItem: vi.fn().mockResolvedValue({ id: '1', name: 'Test' }),
});
const result = await getItemAction.handler({ item_id: '1' });
expect(result.content[0].data).toEqual({ id: '1', name: 'Test' });Hard gate — LLM calls over query results
Do not map chat completions (Atlas agents/chat, OpenAI/Anthropic, useAtlasChat().send) over DMS/SDK rows. Prefer one sendAgentMessage or a resource the sidebar agent can read.
If per-item completions are an explicit product requirement (default: no):
| Rule | Limit |
| Default | 5 completions per user-initiated action |
| Ceiling | 50 — never generate code that can exceed this |
| Cache | space:externalId:lastUpdatedTime; hits do not spend budget |
| Batch | One prompt covering N items, not N calls |
| Trigger | User-initiated only — never on render, poll, or an unbounded list |
| UX | Say when the cap truncated the set |
Forbidden: items.map((row) => complete(row)), Promise.all of completions over a query page.
Checklist
- Topbar Atlas launcher (
use-topbar); no in-app chat widget @cognite/app-sdk@0.3.1+;setApi(() => resolvedApi); catch outside-Fusion rejection- Triggers hidden when
apiis null; server registered/unregistered with.catch() - Resource descriptions say what/when; action names
snake_case; mutating actions require prior approval - Factories take services as args; LLM-over-rows capped (5 / max 50) and cached if present
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

