Skip to content

clawosiris/forge

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

23 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Forge — Multi-Agent Engineering Process for OpenClaw

A structured, multi-agent software development workflow that separates cognitive concerns: creative agents write specs and code, critical agents review them, and adversarial agents break assumptions.

Architecture

Your OpenClaw Instance
  │
  │ "start engineering workflow for project-x: add user auth"
  │
  ├── Supervisor (Forge)          persistent sub-agent, manages pipeline
  │     ├── Analyst / Spec Writer   drafts specs from requirements
  │     ├── Spec Reviewer           validates specs, produces test criteria
  │     ├── Implementer (Codex)     writes code + tests (task-spec driven)
  │     ├── PR Reviewer             reviews code against spec (diff-based)
  │     ├── Code Reviewer           holistic codebase quality review
  │     └── Chaos Agent (Ralph)     adversarial testing, breaks assumptions
  │
  ├── CI/CD Pipeline Monitor       polls CI, auto-diagnoses failures
  ├── GitHub Issue & PR Lifecycle   triage, track, close, review iterations
  └── Human approval gates at spec approval + merge

The supervisor runs as a persistent sub-agent session — your main OpenClaw agent stays general-purpose and spawns workflow supervisors on demand. Multiple workflows can run in parallel.

Quick Start

# Clone
git clone https://github.com/clawosiris/forge.git
cd forge

# Deploy (new standalone instance with container sandboxing)
./deploy.sh

# Or add to an existing instance
./deploy.sh --addon

What's Included

File Purpose
deploy.sh Deployment script with container sandboxing setup
config/openclaw-standalone.json5 Complete config for a fresh instance
config/openclaw-addon.json5 Merge guide for existing instances
workspace/AGENTS.md Main agent instructions (includes workflow routing)
workspace/SOUL.md Main agent persona
workspace/templates/forge-supervisor.md Supervisor spawn template
workspace/templates/agents/ Specialist agent prompt templates
workspace/knowledge/ Project knowledge directory (populate per-project)
docs/deployment-plan.md Full architecture documentation

Workflow Tiers

Not every change needs the full pipeline:

Tier When Agents Used
Small Bug fix, config, <50 LOC Implementer + PR Reviewer
Medium Feature, refactor, 50-500 LOC + Analyst + Spec Reviewer
Large New system, architecture, >500 LOC + Chaos Agent + Code Reviewer (parallel)

Key Features

CI/CD Pipeline Awareness (#1)

The supervisor monitors CI after implementation, auto-diagnoses failures, and spawns targeted fixes (max 3 attempts before escalating). PR review is gated on CI green.

Code Review Agent (#2)

Dedicated holistic code reviewer (separate from PR reviewer) that analyzes the full codebase for quality, architecture, and security — not just the diff.

Codex-First Implementation (#3)

Implementer uses a task-spec-driven pattern: write TASK.md → Codex --full-auto → validate → iterate. This produces the most reliable output.

Spec-Implementation Drift Detection (#4)

After implementation, the supervisor compares what was built against the spec and produces a spec-delta.md flagging scope creep and gaps.

License & Compliance (#5)

Optional compliance stage with SPDX header enforcement, dependency auditing, and SBOM generation. Driven by knowledge/compliance.md.

Test Infrastructure as First-Class Artifact (#6)

Analyst produces test-infrastructure.md alongside the OpenSpec. Mock servers, fixtures, and test doubles get equal review rigor.

Release Management (#7)

Optional release stage: version bump, tag, CHANGELOG, release workflow monitoring — all tracked in workflow-state.json.

Cross-Repository Coordination (#8)

Track downstream impacts across repos. Supervisor creates follow-up issues in dependent repos and coordinates sequential cross-repo work.

Formatting Pre-Commit Gate (#9)

Implementer runs formatter → linter → tests → docs before declaring complete. Eliminates trivial CI failures.

Sandbox-Aware Test Design (#10)

Patterns for feature-gated tests, graceful PermissionDenied skips, and cwd-relative paths for container compatibility.

GitHub Issue & PR Lifecycle (#11)

Full lifecycle management: issue triage, label-based state tracking, PR review iteration handling, and optional cron-based monitoring.

Requirements

  • OpenClaw installed
  • Docker (for container sandboxing)
  • API keys: Anthropic (primary), OpenAI (fallback)
  • At least one messaging channel configured (Signal, Discord, Telegram, etc.)

Documentation

See docs/deployment-plan.md for the full architecture, state machine, failure handling, and observability details.

License

MIT

About

Forge — Multi-agent engineering process for OpenClaw. Structured spec-driven development with cognitive separation: creative agents write, critical agents review, adversarial agents break.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages