Forge for Openclaw

Forge is a repair-inspect loop orchestrator that automates code fixes with independent verification and dependency-aware parallel execution.

melody1015
v1.0.0
Mar 1, 2026
0
1.1k
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install forge-loop

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install forge-loop 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 Forge?

Forge is a sophisticated orchestration skill designed for Openclaw Skills that manages the full lifecycle of automated code repairs. It utilizes a robust state machine to transition tasks from pending to repairing and finally to an independent inspection phase. This ensures that every code change is not just applied, but empirically verified by a separate QA process, preventing the common issue of superficial or broken automated fixes.

By prioritizing safety and reliability, Forge integrates multi-layer guardrails, including protected file lists and pre-commit diff checks. It is built for developers who need to process large volumes of audit findings, bug reports, or review recommendations efficiently without sacrificing codebase integrity.

Forge Use Cases

  • Managing large-scale code repairs from security audits or automated review boards.
  • Orchestrating complex refactoring tasks that require specific execution orders and dependencies.
  • Implementing verified bug fixes where independent QA inspection is required for every change.
  • Automating repetitive code maintenance while ensuring rollback protection and safety guardrails.

How Forge Works

  1. Initialize the session in the project directory to create the state persistence layer.
  2. Add individual repair tasks defining specific success criteria, priority levels, and task dependencies.
  3. Generate an execution plan that organizes tasks into parallel waves based on their dependency graph.
  4. Execute the run command which spawns specialized repair and inspector agents to process the tasks.
  5. The state machine automatically handles failures by looping back to the repair phase for up to 5 attempts.
  6. On successful inspection, the system can automatically commit changes and generate a comprehensive completion report.

Forge Setup

To integrate this skill into your Openclaw Skills environment, follow these steps:

# Navigate to your project directory
cd /path/to/project

# Initialize the forge session
python3 ~/clawd/skills/forge/scripts/forge.py init

# Add tasks with criteria and dependencies
python3 forge.py add "Fix null pointer" --criteria "No crash on empty input" --priority P0

# Execute the workflow
python3 forge.py run

Forge Data Schema & Taxonomy

Forge organizes its operations through a structured set of files and metadata to ensure state recovery and experience accumulation within Openclaw Skills:

File/Directory Description
forge-state.json Stores the current state of all tasks for crash recovery and persistence.
forge-reflections.jsonl Contains project-specific repair patterns and solution templates.
forge-output/ A directory housing specific task definitions and agent results (JSON/MD).
protected-files.txt An optional list of files that agents are strictly forbidden from modifying.
doc-sync-manifest.yaml Maps code authorities to documentation consumers for automatic sync checks.

Forge Advanced Features

  • Dependency-Aware Parallelism: Efficiently executes independent tasks in parallel while respecting sequential constraints.
  • Two-Layer Reflection System: Shares universal repair patterns across projects while maintaining project-specific context.
  • Doc-Sync Checker: Automatically identifies when code changes require corresponding updates in your documentation files.
  • Three-Layer Safety Guardrails: Combines file-level protection, pre-commit diff analysis, and strict prompt constraints to prevent destructive edits.
  • State Persistence: High-reliability orchestration that can recover and resume repair sessions after any system interruption.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*