openforge
Staged, multi-model PRD execution for OpenClaw. Write a PRD with phased sections, model routing, and validation gates — OpenForge executes it across local and cloud models with automatic escalation, scope verification, quality checks, and learning accumulation. Route simple tasks to cheap models, hard tasks to powerful ones, and reviews to premium reasoning.
Install via CLI (Recommended)
clawhub install openclaw/skills/skills/bloodandeath/openforgeWhat This Skill Does
OpenForge is a robust execution engine for OpenClaw designed to transform high-level Product Requirement Documents (PRDs) into actionable, multi-stage development pipelines. By leveraging a sophisticated routing system, OpenForge intelligently delegates sub-tasks to the most appropriate AI models based on complexity and requirements. It automates the entire development lifecycle—from code generation and scaffolding to quality assurance and review—ensuring that simple tasks utilize cost-effective local models while complex, architectural challenges are handled by powerful cloud-based or premium reasoning models. With built-in state management, validation gates, and automatic escalation, OpenForge ensures your projects remain on track, documented, and consistently high-quality.
Installation
OpenForge requires a Unix-like environment with Python 3.11+ and the uv package manager installed. Begin by navigating to the skill directory and executing the installation script: bash <skill_dir>/scripts/install.sh. Ensure that git and openclaw are present in your system path. Once installed, you can verify your environment by running the validation tool against a test PRD file to ensure all routing logic is correctly configured.
Use Cases
- Automated Feature Implementation: Define a complex feature in a PRD, and let OpenForge execute the staging, coding, and testing phases automatically.
- Code Refactoring & Cleanup: Use the 'review-only' PRD template to feed legacy code into the agent, which will systematically suggest and apply improvements based on defined quality checks.
- Multi-Model Orchestration: Standardize your development process by routing boilerplate tasks to local LLMs, saving cloud credits for high-stakes architectural decision-making.
Example Prompts
- "OpenForge, please validate the PRD located at ./docs/auth-refactor.md to ensure all task dependencies and routing rules are correctly defined.
- "Execute the development plan for the user-profile-module.md PRD using the cloud executor for the implementation phase and local models for unit testing.
- "Resume the interrupted execution run with ID 8832-beta and force the inclusion of the recent hotfix branch updates.
Tips & Limitations
- Strict Formatting: OpenForge relies on YAML-block-augmented markdown. Ensure your PRD files strictly follow the structure defined in
references/prd-format.mdto avoid routing errors. - Scope Enforcement: Always define the
produces:field clearly in your tasks. This allows OpenForge to verify that the agent is staying within the project scope. - Escalation: Don't be afraid to set low initial budgets; OpenForge's auto-escalation will promote tasks to more powerful models only if the quality checks fail, saving you significant operational costs.
- Limitations: The tool is designed for code-heavy tasks; it may struggle with highly subjective creative writing or non-code-oriented documentation projects.
Metadata
Not sure this is the right skill?
Describe what you want to build — we'll match you to the best skill from 16,000+ options.
Find the right skillPaste this into your clawhub.json to enable this plugin.
{
"plugins": {
"official-bloodandeath-openforge": {
"enabled": true,
"auto_update": true
}
}
}Tags(AI)
Flags: file-write, file-read, code-execution
Related Skills
design-engineering
Orchestrate iterative design and frontend engineering work through research, planning, sub-agent execution, and validation loops. Use when a visual/UI task requires multiple iterations, when the first implementation needs refinement based on feedback, when choosing between competing technical approaches (Canvas vs SVG vs CSS), or when coordinating sub-agents on design-heavy work. Covers animation architecture decisions, progressive enhancement patterns, performance-aware rendering choices, and the research→plan→execute→validate workflow. Complements frontend-design (which handles aesthetics) by adding engineering discipline, iteration management, and technical decision-making.
agent-cost-monitor
Real-time token usage and cost tracking across all your OpenClaw agents — alerts, budgets, and optimization tips
web-monitor
Monitor web pages for content changes and get alerts. Track URLs, detect updates, view diffs. Use when asked to watch a website, track changes on a page, monitor for new posts/content, set up page change alerts, or check if a site has been updated. Supports CSS selectors for targeted monitoring.
webapp-testing
Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.
code-reviewer
Conduct rigorous, adversarial code reviews with zero tolerance for mediocrity. Default behavior is a single-model adversarial review that identifies security holes, lazy patterns, edge case failures, and bad practices across Python, R, JavaScript/TypeScript, SQL, and front-end code. Supports an optional `--dual` mode for heavier cross-model iterative review when deeper scrutiny is worth the added cost and latency. Use when users ask to "critically review my code", "critically review" code or a PR, "critique my code", "find issues in my code", "find issues" in code, ask "what's wrong with this code", ask to "review this code", "critique my PR", say "double review this", or request a "cross-model review". Scrutinizes error handling, type safety, performance, accessibility, and code quality. Provides structured feedback with severity tiers (Blocking, Required Changes, Suggestions, Noted) and specific, actionable recommendations.