Agent skill

api2cli

Generate a working CLI from any API, then wrap it in a Claude Code skill. Point it at API docs, a live URL, or a peek-api capture and get a dual-mode Commander.js CLI (human + agent output) plus a ready-to-use skill folder. Use when user wants to wrap an API in a CLI, generate a CLI from API docs, turn an API into a command-line tool, scaffold a CLI from discovered endpoints, or create a skill for an API.

Stars 420
Forks 15

Install this agent skill to your Project

npx add-skill https://github.com/alexknowshtml/api2cli/tree/main/skill

SKILL.md

api2cli

Generate a working Node.js CLI from any API, then wrap it in a Claude Code skill. Discovers endpoints, scaffolds a dual-mode Commander.js CLI with a full-featured API client, and creates a skill folder so Claude knows how to use it.

Workflow

  1. Identify the API -- user provides a docs URL, a live API base URL, or a peek-api capture
  2. Discover endpoints -- parse docs, probe the API, or read a peek-api catalog
  3. Build endpoint catalog -- normalize all discovered endpoints into a standard format
  4. Generate CLI -- scaffold Commander.js CLI from the catalog
  5. User chooses destination -- scaffold into current project or create standalone project
  6. Generate skill -- create a SKILL.md that teaches Claude how to use the generated CLI

Step 1: Identify the API

Ask the user:

  • "What API do you want to wrap? Share a docs URL, a base URL, or point me at a peek-api capture."

Determine which discovery paths to use based on what they provide:

Input Discovery Path
Docs URL (e.g., https://docs.stripe.com/api) Docs parsing + active probing
Base URL (e.g., https://api.example.com/v1) Active probing
peek-api capture dir (e.g., ./peek-api-linkedin/) Read existing catalog
Live website URL Suggest running peek-api first, then active probing

Also ask:

  • "What auth does this API use?" (API key, Bearer token, cookies, OAuth, none)
  • "Do you want this CLI in your current project or as a standalone project?"

Step 2: Discover Endpoints

Use all applicable discovery paths. Combine results into a single catalog.

Path A: Docs Parsing

  1. Fetch the docs URL with WebFetch
  2. Extract endpoint information: method, path, description, parameters, request/response examples
  3. Look for pagination patterns, auth requirements, rate limit info
  4. Follow links to sub-pages for individual endpoint docs if the main page is an index

Path B: Active Probing

  1. Check well-known paths for API specs:
    • /.well-known/openapi.json, /.well-known/openapi.yaml
    • /openapi.json, /openapi.yaml, /swagger.json, /swagger.yaml
    • /api-docs, /docs, /api/docs
    • /graphql (with introspection query)
  2. Try OPTIONS on the base URL and common resource paths
  3. Probe common REST patterns: /api/v1/, /api/v2/, /v1/, /v2/
  4. For each discovered resource, try standard CRUD: GET /resources, GET /resources/:id, POST /resources, etc.
  5. Parse response shapes to understand data models
  6. Check response headers for rate limit info (X-RateLimit-*, Retry-After)
  7. Check for pagination patterns in responses (next, cursor, page, offset)

See references/discovery-strategies.md for detailed probing patterns.

Path C: peek-api Capture

  1. Read the capture directory: endpoints.json, auth.json, CAPTURE.md
  2. Parse endpoints into the standard catalog format
  3. Extract auth headers and cookies from auth.json

If peek-api is not installed or no capture exists, tell the user:

To capture endpoints from a live site, install peek-api:
  git clone https://github.com/alexknowshtml/peek-api
  cd peek-api && npm install
  node bin/cli.js https://example.com

Step 3: Build Endpoint Catalog

Normalize all discovered endpoints into this format:

typescript
interface EndpointCatalog {
  service: string;           // e.g., "stripe", "nexudus"
  baseUrl: string;
  auth: {
    type: 'api-key' | 'bearer' | 'cookies' | 'oauth' | 'none';
    headerName?: string;     // e.g., "Authorization", "X-API-Key"
    envVar: string;          // e.g., "STRIPE_API_KEY"
  };
  pagination?: {
    style: 'cursor' | 'offset' | 'page' | 'link-header';
    paramName: string;       // e.g., "starting_after", "offset", "page"
    responseField: string;   // e.g., "has_more", "next", "next_page_url"
  };
  rateLimit?: {
    requests: number;
    window: string;          // e.g., "1m", "1h"
  };
  resources: ResourceGroup[];
}

interface ResourceGroup {
  name: string;              // e.g., "customers", "invoices"
  description: string;
  endpoints: Endpoint[];
}

interface Endpoint {
  method: 'GET' | 'POST' | 'PUT' | 'PATCH' | 'DELETE';
  path: string;              // e.g., "/v1/customers/:id"
  description: string;
  parameters: Parameter[];
  requestBody?: object;      // JSON schema or example
  responseExample?: object;
}

interface Parameter {
  name: string;
  in: 'path' | 'query' | 'header';
  required: boolean;
  type: string;
  description: string;
}

Present the catalog to the user for review before generating:

Found 24 endpoints across 5 resources:
  customers (6 endpoints): list, get, create, update, delete, search
  invoices (5 endpoints): list, get, create, send, void
  ...
Ready to generate the CLI?

Step 4: Generate CLI

Generate a dual-mode CLI using Commander.js. The CLI auto-detects human vs agent output via process.stdout.isTTY.

File Structure

In-project scaffold:

scripts/
  {service}.ts                    # Entry point with shebang
  {service}/
    lib/
      client.ts                   # API client (auth, pagination, retry, caching)
      envelope.ts                 # Agent JSON envelope helpers
    commands/
      {resource}.ts               # One file per resource group

Standalone project:

{service}-cli/
  package.json
  tsconfig.json
  bin/
    {service}.ts                  # Entry point with shebang
  src/
    lib/
      client.ts
      envelope.ts
    commands/
      {resource}.ts

Code Generation Patterns

See these references for the patterns to apply during generation:

  • references/api-client-template.md -- API client class with pagination, retry, rate limiting, caching
  • references/agent-first-patterns.md -- JSON envelope, HATEOAS next_actions, context-safe output, error fix suggestions
  • references/commander-patterns.md -- Commander.js subcommands, global options, interactive prompts, colored output

Key Generation Rules

Entry point ({service}.ts):

  • Shebang: #!/usr/bin/env npx tsx
  • Self-documenting root command (no args → prints full command tree as JSON)
  • Global options: --json (force JSON output), --verbose, --config <path>

API client (lib/client.ts):

  • Constructor takes base URL + auth config
  • Auth from env var (name based on catalog.auth.envVar)
  • Built-in pagination matching the API's pattern
  • Retry with exponential backoff for 5xx and 429 errors
  • Rate limiting based on discovered limits
  • Optional response caching

Envelope helpers (lib/envelope.ts):

typescript
const isAgent = !process.stdout.isTTY;

function respond(command: string, result: any, nextActions: Action[] = []) {
  if (isAgent) {
    console.log(JSON.stringify({ ok: true, command, result, next_actions: nextActions }));
  } else {
    return result; // caller handles human rendering
  }
}

function respondError(command: string, message: string, code: string, fix: string, nextActions: Action[] = []) {
  if (isAgent) {
    console.log(JSON.stringify({ ok: false, command, error: { message, code }, fix, next_actions: nextActions }));
  } else {
    console.error(`Error: ${message}`);
    console.error(`Fix: ${fix}`);
  }
  process.exit(1);
}

Command files (commands/{resource}.ts):

  • One file per resource group
  • Each endpoint becomes a subcommand: mycli customers list, mycli customers get <id>
  • list commands: support --limit, --offset/--cursor, --status (if filterable)
  • get commands: take ID as argument
  • create/update commands: accept --data <json> or individual --field flags
  • Every command includes contextual next_actions for agent mode
  • Errors include fix suggestions

Standalone project extras:

  • package.json with commander, tsx as dependencies, bin field pointing to entry
  • tsconfig.json for TypeScript
  • .env.example with the required env var

Step 5: Verify

After generating the CLI:

  1. Verify it runs: Execute with no args, confirm the self-documenting root works
  2. Test one endpoint: Pick a simple GET endpoint, run it, verify output
  3. Move on to Step 6 to wrap the CLI in a skill

Step 6: Generate Skill

Create a Claude Code skill folder that teaches Claude how to use the generated CLI. This is the final step -- it turns the CLI into something any Claude session can pick up and use without reading the code.

Skill Structure

.claude/skills/{service}/
  SKILL.md                    # Skill instructions

SKILL.md Template

Generate a SKILL.md with this structure:

markdown
---
name: {service}
description: Interact with the {Service} API via CLI. Use when user wants to
  {list of actions based on discovered resources, e.g., "list customers,
  create invoices, check order status"}. Commands: {service} {resource} {action}.
---

# {Service} CLI

CLI wrapper for the {Service} API.

## Setup

Set the `{SERVICE_ENV_VAR}` environment variable:
\`\`\`bash
export {SERVICE_ENV_VAR}=your-api-key-here
\`\`\`

## Commands

{For each resource group, list commands with examples:}

### {Resource}

\`\`\`bash
# List {resources}
npx tsx {path/to/cli}.ts {resource} list

# Get a specific {resource}
npx tsx {path/to/cli}.ts {resource} get <id>

# Create a {resource}
npx tsx {path/to/cli}.ts {resource} create --field value
\`\`\`

## Common Workflows

{Generate 2-3 practical workflows combining multiple commands:}

### Example: {Workflow name}
\`\`\`bash
# Step 1: Find the customer
npx tsx {path/to/cli}.ts customers list --status=active

# Step 2: Get their invoices
npx tsx {path/to/cli}.ts invoices list --customer-id=abc123
\`\`\`

## Agent Usage

When piped, all commands return JSON with `next_actions`:
\`\`\`bash
npx tsx {path/to/cli}.ts {resource} list | cat
\`\`\`

Key Rules for Skill Generation

  1. Description is critical -- include specific trigger phrases and list the actions the CLI supports. This is what Claude reads to decide when to use the skill.
  2. Include real command examples -- use the actual CLI path and real subcommand names from the generated CLI.
  3. Generate practical workflows -- combine multiple commands into realistic multi-step scenarios based on how the API's resources relate to each other.
  4. Keep it lean -- the skill should be a quick reference, not a restatement of --help. Focus on what Claude needs to know that it can't infer.

Tell the User

After generating both the CLI and the skill:

CLI generated at {cli_path}
Skill generated at .claude/skills/{service}/SKILL.md

To use the CLI directly:
  npx tsx {cli_path}                           # See all commands
  npx tsx {cli_path} customers list            # List customers

Claude will now automatically use this skill when you ask about {service}.

Reference Files

  • references/discovery-strategies.md -- Detailed probing patterns, well-known paths, GraphQL introspection, response parsing
  • references/api-client-template.md -- Full API client class with pagination, retry, rate limiting, caching
  • references/agent-first-patterns.md -- Agent JSON envelope, HATEOAS, context-safe output, error handling
  • references/commander-patterns.md -- Commander.js subcommands, nested commands, interactive prompts, colored output, config files, testing

Expand your agent's capabilities with these related and highly-rated skills.

davila7/claude-code-templates

verl-rl-training

Provides guidance for training LLMs with reinforcement learning using verl (Volcano Engine RL). Use when implementing RLHF, GRPO, PPO, or other RL algorithms for LLM post-training at scale with flexible infrastructure backends.

23,776 2,298
Explore
davila7/claude-code-templates

openrlhf-training

High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.

23,776 2,298
Explore
davila7/claude-code-templates

gguf-quantization

GGUF format and llama.cpp quantization for efficient CPU/GPU inference. Use when deploying models on consumer hardware, Apple Silicon, or when needing flexible quantization from 2-8 bit without GPU requirements.

23,776 2,298
Explore
davila7/claude-code-templates

Claude Code Guide

Master guide for using Claude Code effectively. Includes configuration templates, prompting strategies "Thinking" keywords, debugging techniques, and best practices for interacting with the agent.

23,776 2,298
Explore
davila7/claude-code-templates

qdrant-vector-search

High-performance vector similarity search engine for RAG and semantic search. Use when building production RAG systems requiring fast nearest neighbor search, hybrid search with filtering, or scalable vector storage with Rust-powered performance.

23,776 2,298
Explore
davila7/claude-code-templates

behavioral-modes

AI operational modes (brainstorm, implement, debug, review, teach, ship, orchestrate). Use to adapt behavior based on task type.

23,776 2,298
Explore

Didn't find tool you were looking for?

Be as detailed as possible for better results