genkit

Genkit

Use this skill when the main question is "should this Firebase AI feature be a direct client SDK integration or a reusable server-owned Genkit workflow?"

The job is not to dump a long Genkit tutorial, CLI catalog, or Firebase product tour. Frame the current packet, choose one ownership boundary and operating mode, then keep Firebase platform operations and broad framework comparisons routed to adjacent skills.

Read references/intake-packets-and-fallbacks.md before handling mixed or ambiguous requests. Read references/modes-and-routing.md before choosing a primary mode. Read references/deployment-and-runtime-boundaries.md when runtime choice is the real open question. Read references/evals-and-observability.md when the workflow already exists and confidence is the bottleneck.

When to use this skill

• A mobile or web app needs direct Firebase AI Logic client SDK integration, including model setup, structured output, streaming, App Check, or safe client/server boundaries
• A backend or full-stack feature needs a reusable Genkit flow instead of one-off provider calls scattered through route handlers
• The work needs typed contracts, tool calling, retrieval, prompt files, evaluation, tracing, or deployment under a server-owned boundary
• The request is explicitly about Firebase AI Logic, Genkit, or deciding between their client-owned and server-owned shapes
• An existing Firebase AI feature needs debugging, evaluation, observability, or launch-readiness review

When not to use this skill

• The main job is Firebase bootstrap, emulator usage, hosting/functions deploy, auth/login, or admin CLI work → firebase-cli
• The request is generic frontend streaming/rendering with no Firebase AI SDK or Genkit boundary → relevant frontend/web skill
• The real question is broad framework choice (Genkit vs Vercel AI SDK vs other frameworks) → survey
• A plain provider SDK or simple route handler is enough and Firebase/Genkit is not a requirement → note the fallback and keep the answer lightweight

Instructions

Step 1: Frame the current packet

Record the smallest useful intake before recommending Genkit.

Capture:

• app shape: web | mobile | backend | fullstack | mixed | unknown
• ownership: client feature | backend capability | mixed | unknown
• packet: route handler | feature brief | architecture note | deployed flow | eval/trace complaint | deploy plan | none
• workflow need: simple generation | structured output | tools | retrieval | prompt files | evals | observability | deployment | unknown
• delivery pressure: single endpoint | multi-surface reuse | launch readiness | migration | reliability concern | unknown

Quick frame:

App shape: fullstack
Ownership: backend capability
Packet: existing API route + support feature brief
Workflow need: retrieval + one ticket tool + evals later
Delivery pressure: reuse across web app and internal ops panel

Step 2: Choose the intake packet first

Use references/intake-packets-and-fallbacks.md.

Pick the packet the user actually has now:

• new backend capability packet
• existing route/handler packet
• deployed flow quality packet
• deployment/runtime packet
• comparison/fallback packet
• no usable packet yet

Output this step as:

## Intake Packet
- Current packet:

Rule: do not force a server-owned Genkit flow just because the app already uses Firebase.

Step 3: Choose the ownership layer

Make the client-versus-server decision explicit before choosing a mode.

Choose client-ai-logic when the dominant need is:

• direct mobile/web Firebase AI Logic SDK usage
• client streaming or structured output with a small app-owned feature boundary
• App Check, client-visible model configuration, or client/server data-exposure rules

Choose a server-owned Genkit mode when the dominant need is:

• a reusable backend AI contract or typed flow boundary
• one place to own tools, retrieval, prompts, evaluation, or tracing
• a capability shared across clients, jobs, or protected server-side resources

Do not force either layer when a plain provider SDK, ordinary route handler, or broader framework survey is the better fit.

State the decision in one line:

Step 4: Choose one primary operating mode

Pick one primary mode from references/modes-and-routing.md.

Primary modes:

• client-ai-logic
• flow-foundation
• tool-and-agent
• retrieval-and-prompt
• evaluation-and-observability
• deployment-runtime
• comparison-or-fallback

Rule: one primary mode, optional secondary mode. Do not mix client SDK wiring, backend flow design, deployment ops, and architecture comparison into one blob.

Step 5: Freeze the smallest feature boundary

For client-ai-logic, define:

• one app feature and model interaction
• input/output or streaming contract
• client-visible versus server-only data and configuration
• App Check/auth/error handling and abuse boundaries

For a server-owned Genkit mode, define:

• one named backend capability and typed flow contract
• what must remain server-side
• tools/retrieval/prompts, if any
• which clients or jobs call it

Avoid mega-flows, exposing server secrets to clients, or adding tools/retrieval before one basic path works.

Step 6: Name the fallback or route-out honestly

Use references/intake-packets-and-fallbacks.md.

Common route-outs:

• Firebase CLI / emulator / deploy / admin work → firebase-cli
• generic frontend rendering with no Firebase AI SDK boundary → relevant frontend/web skill
• broad framework comparison or architecture uncertainty → survey
• thin non-Firebase model call → plain provider SDK / route-handler fallback
• durability, retries, or background orchestration dominating the problem → queue/job/workflow substrate

Step 7: Pick the smallest next slice

Do not jump to a giant system diagram. Return one slice:

• wire one client SDK generation/streaming path with validation and abuse boundaries
• define or wrap one server flow contract
• add one tool or retrieval boundary
• add one representative eval set
• choose one runtime/deploy shape

Step 8: Use evals and traces when confidence is the bottleneck

Use references/evals-and-observability.md.

When the workflow already exists, prefer:

1. representative inputs
2. local trace review / Developer UI inspection
3. small eval set
4. contract / prompt / tool cleanup
5. rollout only after the evidence loop is good enough

Step 9: Return the Firebase AI implementation brief

Examples

Example 1: Reusable backend support workflow

Input: “Build a Genkit backend flow for our support app: retrieve help articles, call one ticket tool, and expose one server endpoint the web app can reuse.”

Expected shape: tool-and-agent or retrieval-and-prompt, explicit server-owned flow boundary, and one tool/retrieval plan.

Example 2: Direct Firebase app feature

Input: “Add Gemini-powered summaries directly inside our Firebase web app with the Firebase SDK.”

Expected shape: client-ai-logic, with an app-owned SDK boundary, structured output or streaming contract, App Check/auth, and no unnecessary server flow.

Example 3: Existing flow needs confidence before launch

Input: “Our Genkit flows work locally, but we need a practical eval and observability plan before deploying to Cloud Run.”

Expected shape: evaluation-and-observability, small evidence loop, and runtime specifics kept separate from Firebase CLI operations.

Example 4: Framework choice is still unclear

Input: “Should we use Genkit, Firebase AI Logic, Vercel AI SDK, or just direct SDK calls for this Firebase app?”

Expected shape: comparison-or-fallback, route broad framework selection to survey, then return here with either client-ai-logic or a server-owned Genkit mode.

Best practices

1. Choose the client or server ownership boundary before naming tools or models.
2. Prefer one crisp feature/flow boundary over a giant AI feature bucket.
3. Keep secrets and protected resources server-side; define App Check/auth and abuse boundaries for direct client work.
4. Acknowledge plain route-handler / provider-SDK fallbacks when they are enough.
5. Treat runtime choice as an architecture decision, not proof that Genkit is mandatory.
6. Use traces and evals before widening rollout.
7. Sync compact discovery surfaces whenever the front-door boundary changes.

References

• Firebase Genkit docs: https://firebase.google.com/docs/genkit
• Genkit docs: https://genkit.dev/docs/
• Genkit flows docs: https://genkit.dev/docs/js/flows/
• Genkit client access docs: https://genkit.dev/docs/client/
• Firebase AI Logic docs: https://firebase.google.com/docs/ai-logic
• Firebase AI Logic web SDK docs: https://firebase.google.com/docs/ai-logic/get-started?api=web
• ../firebase-cli/SKILL.md
• ../survey/SKILL.md
• ../survey/SKILL.md