swamp（swamp 测试）

Swamp — API Modeling & *Claw Enrichment

You are an AI agent that uses Swamp to model any API, test it, and turn it into a reusable *Claw capability. Swamp is an AI-native automation CLI that represents external resources (APIs, CLIs, cloud services) as typed models with executable methods and composable workflows.

Repository Setup

Before any operation, check if the current directory is a swamp repository. If not, initialize one:

swamp repo init --tool claude

This creates a .swamp/ directory with the necessary structure. Use --tool claude to set up the AI agent integration for Claude/OpenClaw.

API Modeling Workflow

Step 1: Discover Model Types

Find available model types to understand what kinds of resources can be modeled:

swamp model type search
swamp model type describe <type>

The command/shell type is the most versatile — it can model any API via shell commands (curl, httpie, CLI tools).

Step 2: Create a Model

Create a new model definition for the API:

swamp model create <type> <name>

Example: swamp model create command/shell github-issues

Step 3: Edit the Model Definition

Open and edit the model's YAML definition to configure endpoints, authentication, parameters, and methods:

swamp model edit <name>

When editing, define:

• Methods: The operations this API supports (list, create, update, delete)
• Inputs: Parameters each method accepts
• Authentication: Reference vault secrets via CEL expressions
• Commands: The actual shell commands (curl calls, CLI invocations) that execute each method

Step 4: Validate

Ensure the model definition is well-formed:

swamp model validate <name>

Fix any validation errors before proceeding.

Testing & Execution

Run a Method

Execute a method on your model to test it:

swamp model method run <model_name> <method_name>

Inspect Results

Check outputs, logs, and data from the execution:

swamp model output get <output_id_or_model_name>
swamp model output logs <output_id>
swamp model output data <output_id>

Review History

See past executions:

swamp model method history search
swamp model method history get <output_id_or_model_name>

Iterate on the model definition until methods return the expected results.

Workflow Orchestration

Chain multiple model methods into automated workflows with dependency ordering:

swamp workflow create <name>
swamp workflow edit <name>
swamp workflow validate <name>
swamp workflow run <name>

Workflows support:

• Parallel job execution
• Dependency ordering between jobs
• Trigger conditions via CEL expressions
• Cross-model references (one model's output feeds into another)

Check workflow results:

swamp workflow history get <name>
swamp workflow history logs <run_id>

Vault & Credentials

Store API keys and secrets securely — never hardcode them:

swamp vault create <type> <name>
swamp vault put <vault_name> <KEY=value>
swamp vault list-keys <vault_name>

Reference secrets in model definitions using CEL expressions like vault.get("my-vault", "API_KEY").

Data Management

Inspect outputs and artifacts across models:

swamp data list <model_name>
swamp data get <model_name> <data_name>
swamp data search <query>

Swamp-Club Authentication

Swamp-club is the community registry and collaboration layer. Authenticate to push/pull extensions and share work:

swamp auth login
swamp auth login --server <url>
swamp auth whoami
swamp auth logout

• swamp auth login opens a browser-based login flow by default
• Use --no-browser with --username and --password for headless/CI environments
• Set SWAMP_CLUB_URL env var to target a custom server
• Always verify identity with swamp auth whoami after login

Authentication is required before pushing extensions to the registry.

Extensions

Extensions expand Swamp's model types and capabilities. Use the extension registry to share and reuse custom models:

Pull an Extension

Install a community or team extension from the registry:

swamp extension pull <extension_name>
swamp extension pull <extension_name> --force

Use --force to overwrite existing files without prompting.

List Installed Extensions

See what extensions are currently installed:

swamp extension list

Push an Extension

Publish your own extension to the swamp registry (requires swamp auth login first):

swamp extension push <manifest-path>
swamp extension push <manifest-path> --dry-run

Use --dry-run to build the archive locally and verify it without actually publishing. Use -y to skip confirmation prompts.

Remove an Extension

Uninstall a pulled extension and clean up its files:

swamp extension remove <extension_name>

Extension Workflow

The typical flow for extending Swamp with new model types:

1. Create custom TypeScript model definitions in extensions/models/
2. Test locally with swamp model create <your-type> <name>
3. Package with a manifest file
4. Dry-run: swamp extension push ./manifest.yaml --dry-run
5. Publish: swamp extension push ./manifest.yaml
6. Others install: swamp extension pull <your-extension>

Enriching *Claw with New Capabilities

Once a Swamp model is validated and working, turn it into a standalone *Claw skill:

1. Export the model: Use swamp model get <name> --json to extract the full definition
2. Generate a SKILL.md: Create a new skill file that wraps the Swamp model's methods as agent instructions
3. Include setup instructions: Document the required env vars, binaries, and vault configuration
4. Publish to ClawHub: Share the skill with the community via clawhub publish

The generated skill should:

• Require swamp in its bins dependency
• Reference the swamp repo and model by name
• Map each model method to a clear agent instruction
• Include examples of typical invocations

Sharing via Extensions

For reusable model types (not just individual models), publish as a Swamp extension:

1. Build the custom model type in extensions/models/
2. Push to the Swamp registry: swamp extension push ./manifest.yaml
3. Create a *Claw skill that pulls the extension and uses it: swamp extension pull <name>

This creates a two-layer sharing model: extensions for model types (Swamp registry), skills for agent workflows (ClawHub).

Examples

Model a REST API

"Model the JSONPlaceholder API so I can list and create posts"

1. swamp model create command/shell jsonplaceholder
2. Edit to add methods: list-posts (GET /posts), create-post (POST /posts)
3. swamp model method run jsonplaceholder list-posts
4. Verify output, iterate

Set Up a Weather Integration

"Create a weather model that fetches forecasts by city"

1. swamp model create command/shell weather
2. Edit to add a forecast method using curl -s "wttr.in/{city}?format=j1"
3. Test with swamp model method run weather forecast

Chain APIs in a Workflow

"Create a workflow that fetches GitHub issues, summarizes them, and posts to Slack"

1. Create models: github-issues, slack-webhook
2. Create workflow: swamp workflow create issue-digest
3. Define jobs with dependencies: fetch issues -> format summary -> post to Slack
4. swamp workflow run issue-digest

Graduate to a *Claw Skill

"Turn my working weather model into a skill others can install"

1. swamp model get weather --json to export
2. Create a new SKILL.md wrapping the weather model commands
3. clawhub publish ./weather-skill

Share a Custom Extension

"Publish my custom Stripe model type so the team can use it"

1. swamp auth login to authenticate with swamp-club
2. swamp extension push ./stripe-models/manifest.yaml --dry-run to verify
3. swamp extension push ./stripe-models/manifest.yaml to publish
4. Team members install: swamp extension pull stripe-models

Important Notes

• Always use --json flag when you need to parse Swamp output programmatically
• Use swamp model validate before running to catch definition errors early
• Store all credentials in vaults, never in model definitions
• Use -v (verbose) flag when debugging unexpected behavior
• All Swamp operations run locally — credentials stay on the user's machine
• Authenticate with swamp auth login before pushing extensions to the registry
• Use swamp extension push --dry-run to verify an extension before publishing
• Use swamp auth whoami to confirm your identity before registry operations