A structured methodology to replicate an AI agent configuration across multiple clients using a single context source and a strict, validation-gated generator.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-deployment-kit
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agent-deployment-kit using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Agent Deployment Kit?
The Agent Deployment Kit offers a structured approach to deploying a single AI agent architecture across multiple clients while helping prevent configuration drift. Instead of copying folders manually, this approach uses one template skeleton combined with client-specific context files. By utilizing Openclaw Skills to establish this framework, developers maintain a single source of truth for prompts, safety rules, and operating procedures.
At the heart of the methodology is a local generator that translates questionnaire variables into concrete deployments while refusing unknown or incomplete values. This design mitigates the risks of configuration drift, residue leaks where one client's data accidentally transfers to another, and bypassed safety constraints.
Agent Deployment Kit Use Cases
- Scaling Deployments to Multiple Clients: When you need to ship the same underlying AI agent architecture to N clients, ensuring each gets their specific branding, integrations, and tools without diverging the base code.
- Updating Configurations Safely: Modifying a setting (like a booking link or key parameter) in one centralized questionnaire instead of directly editing live files.
- De-scoping Agent Features: Cleanly stripping out-of-scope workflows (such as crons and playbooks) for clients who did not purchase specific features, avoiding agent improvisation on dead code.
- Compliance and Safety Verification: Ensuring all deployed agents run with explicit data retention rules, clear legal bases, and honest AI disclosure markers.
How Agent Deployment Kit Works
- Define the Context: Developers populate
context/QUESTIONNAIRE.jsonwith client-specific variables (with no secrets) and define key lists incontext/secrets.example.json. - Isolate Secrets: Real API keys and credentials are saved locally in
context/secrets.local.json, which is gitignored to avoid code repository leaks. - Build the Skeleton: The core agent files (such as prompt files, playbooks, health checks) are designed with placeholders using the
{{PLACEHOLDER}}format within theskeleton/directory. - Run the Generator: A local script processes the templates and outputs final deployment files to the disposable
out/directory, checking against sentinel patterns to block incomplete configurations. - Deploy with Safety Gates: The deployment process verifies folder ownership, creates secure backups, applies system permissions (like 0600 for secrets files), and performs post-render syntax and keyword validations.
Agent Deployment Kit Setup
To begin organizing your agent deployment with this methodology:
- Ensure you have the required runtime dependencies installed (such as Node.js and grep).
- Create a new kit structure and initialize the context files based on the structure recommended in Openclaw Skills:
mkdir -p agent-kit-myclient/{context,skeleton,generator,out}
cd agent-kit-myclient
- Configure your local
.gitignoreto prevent secret leaks and backup issues:
cat <<EOT >> .gitignore
context/secrets.local.json
out/
*.bak-*
EOT
- Author your custom generator script (e.g.,
generator/generate.mjs) to process templates while incorporating strict validation patterns for all input variables.
Agent Deployment Kit Data Schema & Taxonomy
The file and data organization ensures clean separation between base logic and client configuration:
| Directory / File | Role | Version Control |
|---|---|---|
context/QUESTIONNAIRE.json |
Client-specific, non-secret parameters | Committed (ASCII only) |
context/secrets.local.json |
Active production secrets and credentials | Gitignored |
skeleton/ |
Variable-based template engine directory | Committed |
out/ |
Compiled deployment-ready files | Gitignored (Rebuilt on run) |
generator/generate.mjs |
The custom parser and strict validation engine | Committed |
Questionnaire Structure
Your QUESTIONNAIRE.json must be structured into exactly two main blocks:
_meta: Human-readable description of metadata and scope (not rendered).config: Machine-facing properties (such as installation directories and model identifiers).context: Plain-text parameters rendered into the agent's prompts and behaviors.
Agent Deployment Kit Advanced Features
- Dual-Mode Sentinel Gate: Supports a default warning mode for scoping stages and a strict deployment mode that fails immediately on empty parameters or unresolved placeholder strings.
- Static-Body Guardrails: Load-bearing instructions (like AI honesty markers and data retention policies) are hard-coded in the template layout, with context files only modifying basic parameters.
- Residue Hunt Audits: Automated grep-based workflows to search finalized model contexts for leftover generic identifiers, staging URLs, or third-party handles.
- Field of Vision Duplication: Replicates core safety rules inside
AGENTS.mdandTOOLS.mdto guarantee sub-agents respect limits even when they cannot access the main prompt.
SKILL.md
Loading
Related Openclaw Skills
Canvora Visual Generation Skill
Canvora is a CLI-driven utility that converts raw text, web URLs, and documents into high-quality, on-brand marketing visuals and structured slide decks.
Tristan RFQ Overseer
Tristan RFQ Overseer turns inbound RFQs into priced, compared, and approval-gated quotation drafts with vault-backed state.
UI Master Static
Openclaw Skills guidance for building production-grade interfaces with native CSS, BEM, modern layouts, and polished motion.
GIF Picker
GIF Picker lets Openclaw agents choose a local reaction GIF by tag and return a reliable `MEDIA:` path for chat replies.
Agent Handshake
Openclaw Skills turns AI servers into discoverable, persistent agents with one-command registration, acceptance, and handshake workflows.
Novelty Validator
An academic research verification tool that extracts claims from meeting records and validates novelty against a 1.2-billion-document database.
Featured*
