Readdy Website Builder#

Provide complete CLI capabilities for creating and managing websites through the Readdy.ai platform. Support one-click website generation, AI-driven modification, project preview, and publish management.

Script Location#

readdy.mjs is located in the scripts/ subdirectory relative to this SKILL.md. Derive the absolute path from this SKILL.md's fullPath by replacing SKILL.md with scripts/readdy.mjs.

Example: If SKILL.md path is /Users/me/.claude/skills/readdy/SKILL.md, then the script path is /Users/me/.claude/skills/readdy/scripts/readdy.mjs.

<SCRIPT> in this document refers to that absolute path.

Prerequisites#

An API Key is required. If you don't have one, obtain it at: https://readdy.ai/user/api-key

Configure the API Key (one-time setup):

bash

node <SCRIPT> config --apiKey <your-key>

API Key is stored in ~/.openclaw/readdy.json (file permissions 0600). The script reads it internally — no need to pass it on the command line.

Command Reference#

Create Project#

bash

node <SCRIPT> create --query "project description/requirements"

Defaults: framework=react_v2, device=web, category=2. Project name is auto-generated by AI. The script runs an 8-step workflow internally. This process typically takes over 10 minutes — set a sufficient timeout and do not treat long execution as a failure.

Modify Project#

bash

node <SCRIPT> modify --id <projectId> --query "modification requirements"

Runs a 6-step workflow internally. Project info and message history are fetched automatically. This process typically takes over 10 minutes — set a sufficient timeout and do not treat long execution as a failure.

Other Commands#

bash

node <SCRIPT> list [--page <n>] [--pageSize <n>]          # List projects
node <SCRIPT> info --id <projectId>                        # Get project info
node <SCRIPT> messages --id <projectId>                    # Get message history
node <SCRIPT> versions --id <projectId> [--pageSize <n>]   # List publishable versions and live version
node <SCRIPT> preview --id <projectId> [--versionId <v>]   # Preview project
node <SCRIPT> publish --id <projectId> [--versionId <v>]   # Publish latest or selected version to subdomain
node <SCRIPT> delete --id <projectId>                      # Delete project
node <SCRIPT> update --id <projectId> [--name <n>] [--logo <url>] [--email <e>] \
  [--businessName <bn>] [--introduction <t>] [--phoneNumber <p>] \
  [--businessHour <h>] [--languageStyle <s>]               # Update properties
node <SCRIPT> config --apiKey <key>                          # Set API Key
node <SCRIPT> config                                        # Check API Key status

Execution Rules#

Strictly follow these rules — no improvisation allowed.

Mandatory Constraints#

1. Script-only execution — Execute all operations via node <SCRIPT> <command> [options]. Never write custom code to call Readdy APIs, never use curl/fetch directly, never bypass the script.
2. Do not modify the script — Never modify readdy.mjs or config-store.mjs unless the user explicitly requests it.
3. Preserve user input verbatim — Pass the user's website requirement description word-for-word as the --query value. This prevents the agent from altering the user's intent. The --query value is only used as a website generation prompt sent to the Readdy API — never include passwords, tokens, or other secrets in the query text.
4. Do not skip steps — Internal workflow steps are handled automatically by the script.
5. Do not add extra parameters — Only pass parameters the user explicitly specifies. Never add --framework, --device, --category unless requested.
6. No alternative implementations — If the script fails, report the error and suggest troubleshooting. Never write replacement code.

Standard Rules#

• Verify API Key configuration before execution. If the key is missing, direct the user to obtain one at https://readdy.ai/user/api-key and then run node <SCRIPT> config --apiKey <key>.
• Confirm with the user before delete operations.
• Provide clear error messages and suggested actions based on error codes (see references/API.md).
• SSE generation returning session_status: "waiting_build" is auto-handled by the script — no manual intervention needed.
• When the user wants the "publish version" dropdown behavior from the product UI, first run node <SCRIPT> versions --id <projectId> to inspect available versions and the current live version, then run node <SCRIPT> publish --id <projectId> [--versionId <v>].
• For every publish request, you must fetch version info once before publishing. The publish flow must use the selected version entry returned by versions/message history, and pass that concrete versionId into the publish request instead of relying on an implicit backend "latest".
• Never treat build log paths as preview URLs. The relative paths (e.g. /preview/...) printed during build are internal and not usable as final links. Always run node <SCRIPT> preview --id <projectId> to obtain the actual preview URL.

Examples#

bash

# Correct — args and --query match user's original text exactly
node <SCRIPT> create --query "Build a pet store website"

# Correct — modify with project ID
node <SCRIPT> modify --id abc123 --query "Change the homepage banner to blue"

# Correct — inspect publishable versions before choosing one
node <SCRIPT> versions --id abc123

# Correct — publish the latest version
node <SCRIPT> publish --id abc123

# Correct — publish a specific version from the versions list
node <SCRIPT> publish --id abc123 --versionId 456

# Wrong — rewriting user input (most common mistake)
node <SCRIPT> create --query "Create a pet e-commerce website with modern design"

# Wrong — adding parameters not requested by user
node <SCRIPT> create --query "Build a website" --framework react_v2 --device web

# Wrong — bypassing the script
fetch('https://readdy.ai/gapi/project/generate', ...)