Api Design First logo

Api Design First

Community
peterbamuhigire
api-design-first

Use when designing HTTP APIs with OpenAPI, versioning, authentication, rate limits, idempotency, error contracts, observability requirements, and choosing REST, webhooks, SSE, or WebSocket delivery.

Overview

Publisherpeterbamuhigire
Repositorychwezi-dev-engine
Skill nameapi-design-first
Stars
29
Forks
6
Bundled files
24
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 24 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by peterbamuhigire on GitHub. Read the source before you install it.

Installation

Install the Api Design First AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/peterbamuhigire/chwezi-dev-engine/tree/main/skills/architecture/api-design-first
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/peterbamuhigire/chwezi-dev-engine.git /tmp/chwezi-dev-engine
mkdir -p .claude/skills
cp -r /tmp/chwezi-dev-engine/skills/architecture/api-design-first .claude/skills/api-design-first
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Api Design First in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Api Design First on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Api Design First is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

API Design First

Acknowledgement: Shared by Peter Bamuhigire, techguypeter.com, +256 784 464178.

Design APIs as contracts before code. This skill produces the OpenAPI 3.1 contract and its companion auth / error / idempotency / observability artifacts that downstream frontend, mobile, SDK, security, and reliability skills depend on.

Use When

  • Designing a new HTTP API (REST or GraphQL decision open), or adding endpoints to an existing API.
  • Normalising an existing API against the house contract (envelope, error model, versioning, rate limits).
  • Producing the OpenAPI contract that SDK, frontend, mobile, contract-testing, and security skills will consume.
  • Choosing an auth method and writing the auth/role matrix for a new service.
  • Defining idempotency keys and observability notes handed to reliability-engineering and observability-monitoring.

Do Not Use When

  • The task is purely client-side consumption of a third-party API (load the relevant SDK or integration skill).
  • The task is full threat modelling — load vibe-security-skill; this skill only records the auth model.
  • GraphQL-specific hardening is needed — load graphql-patterns.

Required Inputs

  • Context map and critical-flow table from system-architecture-design.
  • Access-pattern list from database-design-engineering.
  • Consumers, trust boundaries, and latency budget for each endpoint.
  • Tenancy model (single-tenant, shared schema, schema-per-tenant) so tenant_id propagation can be proven.

Workflow

  • Read this SKILL.md, then load only the references needed for the task (for example openapi-workflow.md for a greenfield design; auth-and-security.md for an auth choice; rest-conventions.md for URL / status-code review).
  • Apply the six-step design workflow before writing any code.
  • Produce the OpenAPI contract in the template format from skill-composition-standards/references/openapi-contract.md.
  • Record the auth model, error model, idempotency map, and observability notes alongside the spec.
  • Run the implementation checklist before merging.

Quality Standards

  • Spec is the source of truth; code follows the spec, never the reverse.
  • Every endpoint has documented auth, error codes, rate limits, pagination where relevant, and observability notes.
  • Tenant scope is derived from the auth token, never from a request body or query string.
  • Breaking changes require a new major version; existing versions are frozen except for bug fixes.
  • British English, no emojis, language-tagged code fences.

Anti-Patterns

  • Writing code first and "generating" the spec afterwards — the spec loses its role as contract.
  • Accepting tenant_id from the request body — enables cross-tenant data leaks.
  • Returning 403 for wrong-tenant lookups — leaks existence; return 404.
  • Wildcard Access-Control-Allow-Origin: * with credentials — browsers ignore, auth breaks, security hole opens.
  • Storing API keys in plaintext — a database leak becomes an immediate credential leak.
  • Mutating an existing version's response schema — every client with hard-coded parsers breaks.

Outputs

  • OpenAPI 3.1 contract covering every endpoint, schema, error, rate-limit header, and example.
  • Auth model: selected method, role/scope matrix, token lifetime, revocation strategy.
  • Error model conforming to the standard envelope with documented error codes.
  • Idempotency map: which endpoints require Idempotency-Key, scope, TTL.
  • Observability notes: per-endpoint span names, SLO class, log fields, audit events.
  • Implementation checklist result.

Evidence Produced

CategoryArtifactFormatExample
CorrectnessOpenAPI contractMarkdown doc plus OpenAPI YAML covering paths, schemas, and error responsesdocs/api/openapi-checkout.yml
CorrectnessContract test resultsCI log or recorded report from API contract tests against the OpenAPI specdocs/api/contract-tests-2026-04-16.md

References

Prerequisites

Load these skills first:

  • system-architecture-design — produces the context map and critical-flow table consumed below.
  • database-design-engineering — produces the access-pattern list that shapes resource design.
  • world-class-engineering — the shared release gate this skill ships into.

Load alongside:

  • vibe-security-skill for threat modelling of the auth model produced here.
  • graphql-patterns whenever building or auditing a GraphQL API (it owns the GraphQL hardening guidance).

When this skill applies

  • A new service is being designed and needs its HTTP surface defined.
  • A feature adds endpoints to an existing API and must follow the house contract.
  • An existing API is being normalised for versioning, error envelope, or rate limits.
  • A partner integration requires a published OpenAPI contract.
  • Mobile or SDK work is blocked on a stable contract for code generation.

Inputs

ArtifactProduced byRequired?Why
Context mapsystem-architecture-designrequiredDefines service/module boundaries the API sits inside
Critical-flow tablesystem-architecture-designrequiredSupplies latency/availability targets for endpoints
Access-pattern listdatabase-design-engineeringrequiredRead/write shapes drive resource model and pagination choices
Failure-mode listsystem-architecture-designoptionalInforms degradation behaviour on dependency outages
Threat modelvibe-security-skilloptionalInforms auth, rate-limit, and abuse-case rules
Tenancy modeldatabase-design-engineeringrequiredProves tenant isolation at the API boundary

Outputs

ArtifactConsumed byTemplate
OpenAPI 3.1 contractfrontend, mobile, SDK, advanced-testing-strategy (contract tests)skill-composition-standards/references/openapi-contract.md
Error modelfrontend, mobile, advanced-testing-strategyskill-composition-standards/references/error-model.md
Auth modelvibe-security-skill, mobile-rbac, ios-rbacinline in spec (components.securitySchemes) + role/scope matrix in auth-and-security.md
Idempotency mapreliability-engineering, stripe-paymentsinline table — see references/auth-and-security.md
Observability notesobservability-monitoringinline table — see references/implementation-checklist.md

Non-negotiables

  • OpenAPI spec is written before a single endpoint is implemented.
  • Tenant scope comes from the auth token, never from the request body or query string.
  • Every response carries the mandatory security headers from auth-and-security.md.
  • Every POST that causes an external side effect (payment, email, webhook, inventory movement) requires an Idempotency-Key.
  • Every new endpoint is added to the observability notes table before merge.
  • Breaking changes never mutate an existing version; they ship as a new major version.

Decision rules

Auth-method selection

Caller shapeChooseFailure mode if wrong
Server-to-server, trusted partner, long-livedAPI Key (SHA-256 hashed)JWT: can't be revoked fast; raw-key storage: leak = instant compromise
Web app with user consent, third-party accessedOAuth2 Authorization CodeImplicit flow: token leaks via URL history; ROPC: phishing vector
Service-to-service with scopes, standards-compliantOAuth2 Client CredentialsAPI key: no standardised scopes or introspection
First-party mobile app / SPAJWT access + refresh (short)Long-lived JWT: revocation impossible; session cookie: breaks on iOS
Legacy / tightly-controlled client with no user flowOAuth2 ROPC (last resort)Credentials pass through the API surface — avoid when any alternative

Versioning action

ChangeActionFailure mode if wrong
Adding optional field, endpoint, or enum valueShip in current version, mark non-breakingSpawning a new version for a safe change: client churn
Removing or renaming a fieldShip as new major version, sunset oldMutating existing version: silent parsing failures in clients
Changing response shape or required paramsShip as new major version, sunset oldAs above
Bug fix that narrows behaviour clients may depend onShip behind a header flag, deprecate quietlySilent fix: clients relying on the bug break in production

Idempotency requirement

Endpoint classRequire Idempotency-Key?TTLFailure mode if skipped
Payment, refund, chargeyes24 hDouble charges on retry
Order/invoice creationyes24 hDuplicate orders on network retry
Outbound email / webhook dispatchyes7 dDuplicate user-visible side effects
Inventory movementyes24 hPhantom stock changes
Read-only (GET, HEAD)non/an/a
Partial update without external effectoptional1 hBenign — but worth enabling on sensitive ops

Cache-Control choice

Resource shapeHeaderFailure mode if wrong
Public, rarely changes (prices)public, max-age=3600no-store: wastes bandwidth, slow clients
Per-user privateprivate, max-age=300public: leaks via shared CDN cache
Sensitive or mutable (auth, cart)no-storepublic: security incident

GraphQL vs REST

SituationChoose
Multiple clients with diverging field needsGraphQL
Public API, webhooks, file uploadsREST
Simple CRUDREST
Real-time subscriptionsGraphQL subscriptions

Failure mode if wrong: GraphQL-first for simple CRUD buys query-complexity DoS risk with no field-flexibility payoff; REST-first for multi-client heavy-reads pushes clients into over-fetch and N+1 round trips.

This table is the REST/GraphQL shortcut only. For real-time, service-to-service, event, or callback interactions (SSE, WebSocket, gRPC, webhooks, broker messaging), decide per interaction with references/api-style-selection.md; one-way live updates rarely need GraphQL subscriptions or WebSocket.

Design workflow (six steps)

  1. Define consumers, latency expectations, and trust boundaries.
  2. Write the API action plan, then model resources and actions around business concepts, not controller names.
  3. Write the OpenAPI contract, including auth, validation, errors, and pagination.
  4. Prove tenancy, authorisation, and idempotency rules before implementation.
  5. Design observability: request IDs, audit events, deprecation path, rate-limit telemetry.
  6. Validate that the API can evolve without breaking current consumers.

Full spec skeleton and per-step detail in references/openapi-workflow.md. For partner, public, long-lived, or workflow-heavy APIs, also load references/practical-api-architecture.md. For consumer ergonomics, endpoint naming, pagination, embedding, versioning, and test expectations, load references/api-you-wont-hate-rules.md.

Response envelope

The canonical envelope produced by this skill — every downstream client depends on it:

json
{
  "success": true,
  "data": { "id": 123 },
  "meta": { "page": 1, "per_page": 25, "total": 142 }
}
json
{
  "success": false,
  "error": {
    "code": "VALIDATION_ERROR",
    "message": "The amount field is required.",
    "documentation_url": "https://api.example.com/errors#VALIDATION_ERROR",
    "fields": { "amount": ["Required"] }
  }
}

Detail, status-code mapping, and pagination mechanics live in references/rest-conventions.md.

Anti-patterns

  • Spec-after-code. The team writes endpoints and exports the OpenAPI spec from code annotations. Fix: draft the spec in a PR first, get sign-off from at least one consumer, then implement.
  • Tenant from body. POST /api/v1/invoices { "franchise_id": 42, "amount": 100 }. Fix: derive franchise_id from the authenticated token server-side; ignore any tenant field on the wire.
  • 403 on wrong-tenant. GET /api/v1/invoices/999 returns 403 Forbidden when invoice 999 belongs to another tenant. Fix: return 404 Not Found — 403 leaks that the resource exists.
  • Wildcard CORS with credentials. Access-Control-Allow-Origin: * plus Access-Control-Allow-Credentials: true. Fix: echo only origins from an allow-list; set credentials true only for those origins.
  • Raw API key storage. INSERT INTO api_keys (key) VALUES ('sk_live_abc123'). Fix: store key_hash = SHA-256(key), return the raw key once on creation, never again.
  • Unversioned breaking change. The team removes a field from GET /api/v1/invoices/{id} because nobody "should" be using it. Fix: ship as v2, keep v1 unchanged, emit Deprecation + Sunset headers on v1.
  • No idempotency on payments. POST /api/v1/payments charges the card. Network retry charges it again. Fix: require Idempotency-Key header, store (tenant, user, key) -> response for 24 h, replay the stored response on duplicate.
  • Unbounded page size. GET /api/v1/invoices?per_page=100000. Fix: clamp per_page to a hard maximum (e.g. 100) server-side.

Read next

  • skill-composition-standards — normalisation standard this skill was rewritten against; templates for the OpenAPI contract and error model live there.
  • observability-monitoring — consumes the observability notes table to build SLOs, alerts, and runbooks.
  • reliability-engineering — consumes the idempotency map for retry and circuit-breaker design.
  • deployment-release-engineering — consumes the versioning plan for migration choreography and rollback windows.
  • vibe-security-skill — consumes the auth model to build the threat model and abuse-case list.

References

Capability contract

Default to contract design and validation. Write specifications or generate clients only when authorised; do not deploy gateways, rotate credentials, or change production routes without approval.

Degraded mode

If consumers, traffic evidence, or the current contract are unavailable, produce a read-only draft with assumptions and unresolved compatibility risks; do not claim implementation readiness.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Api Design First AI skill do?

Use when designing HTTP APIs with OpenAPI, versioning, authentication, rate limits, idempotency, error contracts, observability requirements, and choosing REST, webhooks, SSE, or WebSocket delivery.

Why use Api Design First on TypingMind?

Because you install it once and use it with any model. Api Design First is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Api Design First in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/peterbamuhigire/chwezi-dev-engine/tree/main/skills/architecture/api-design-first. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Api Design First?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Api Design First?

As many as you like. As long as a model supports skills, you can use Api Design First with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Api Design First AI skill free?

Yes. It is published on GitHub by peterbamuhigire under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇