By kenhuangus
Enforce a cross-platform compound engineering loop across Claude Code, Codex, and other agent-managed projects.
Claim files before editing so Claude Code, Codex, and other agents do not collide.
Finish a task by writing durable compound learning and checking the completion gate.
Show active cross-agent ownership claims for conflict avoidance.
Run the required two-round Codex/Claude cross-tool review protocol and record each stage.
Check planning contracts, two-round review artifacts, verification, README freshness, and compound learning before handoff.
Plans architecture and decomposition for compound engineering tasks.
Reviews Codex-authored or other-agent-authored changes from the Claude Code side and checks for ownership conflicts.
Reviews diffs for correctness, security, regressions, tests, and compound-learning opportunities.
Designs and runs focused verification for a scoped compound engineering task.
Use when Codex reviews Claude Code-authored or other-agent-authored changes and verifies ownership claims before integration.
Use when setting up or enforcing a compound engineering loop across coding, writing, research, or documentation projects with Codex, Claude Code, and other agent runtimes.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Compound Orchestrator is a reusable plugin and project bootstrapper for making agent-assisted work compound over time. It installs a shared harness for coding, writing, research, documentation, and mixed projects so every meaningful task leaves behind a plan, review, verification trail, ownership record, README update when relevant, and durable learning for the next task.
It works across Windows, macOS, Linux, and WSL. Claude Code and Codex get first-class support, and any other agent runtime can participate by using the same generated CLI, ownership file, review rubric, and cross-platform verifier.
README.md maintenance policy for keeping project docs current instead of stale.prd.html, planning.html, spec.html, test-cases.html, architecture.html, and users.html.architecture.excalidraw seed referenced from architecture.html.AGENTS.md and CLAUDE.md managed protocol blocks for shared agent behavior.scripts/compound_orchestrator.py, scripts/verify.py, scripts/verify.ps1, and scripts/verify.sh.Use this plugin for:
The core loop is:
brainstorm -> plan -> six planning contracts -> two-round review -> work -> two-round review -> compound -> repeat
The final step is the point. Each completed task should improve the next task by recording a reusable pattern, decision, failure note, review, or compound note.
From this plugin directory on Windows PowerShell:
$py = "python"
& $py .\scripts\compound_orchestrator.py init --target "C:\path\to\project"
& $py .\scripts\compound_orchestrator.py check --target "C:\path\to\project"
If python is not configured on Windows, point $py at a virtualenv, system Python, or bundled agent runtime Python.
From macOS, Linux, or WSL:
python3 scripts/compound_orchestrator.py init --target /path/to/project
python3 scripts/compound_orchestrator.py check --target /path/to/project
The initializer is conservative. It creates missing files, appends managed blocks to existing README.md, AGENTS.md, and CLAUDE.md, and preserves local content unless --force is used for generated files.
After initialization, the target project is portable. Agents can run the copied project-local CLI:
python3 scripts/compound_orchestrator.py check --target .
python3 scripts/verify.py
README.md is part of the product. Update it whenever setup, usage, architecture, workflow, commands, generated outputs, public behavior, manuscript structure, export process, or review workflow changes.
Every README update should:
The plugin installs this rule into generated projects through both the managed README block and .agent-loop/readme-maintenance.md.
Start a task:
python3 scripts/compound_orchestrator.py start --target . --title "Add billing webhook retry handling"
Start a team run:
python3 scripts/compound_orchestrator.py team-start --target . --title "Refactor checkout flow" --mode codex-parallel-agents
Substantial projects should complete the generated planning contracts before implementation:
prd.html -> planning.html -> spec.html -> test-cases.html -> architecture.html -> users.html
Planning should still be parallel where it is independent. Draft PRD, users, architecture, planning, and early test strategy in parallel; then serialize integration, spec.html, test-cases.html, review, and final acceptance.
Claim files or artifacts before editing:
python3 scripts/compound_orchestrator.py claim --target . --tool codex --agent codex-worker --task-id TASK --paths src/payment.py --intent "Implement retry handling"
npx claudepluginhub kenhuangus/compound-orchestratorCookbook Memory for Claude Code: persistent memory (recall / remember) so the agent remembers across sessions. Routes through the Orchestrator; fail-open.
Harness-native ECC plugin for engineering teams - 67 agents, 277 skills, 93 legacy command shims, reusable hooks, rules, MCP conventions, and operator workflows for Claude Code plus adjacent agent harnesses
Claude harness - A harness for solo developers (Vibecoders) to handle full-cycle contract development.
v9.44.1 — Patch release for Gemini environment/version detection and qwen auth gating. Run /octo:setup.
Persistent file-based planning for AI coding agents. Crash-proof markdown plans (task_plan.md, findings.md, progress.md) that survive context loss and /clear, with an opt-in completion gate and multi-agent shared state. Manus-style. Works with Claude Code, Codex CLI, Cursor, Kiro, OpenCode and 60+ agents via the SKILL.md standard. Includes Arabic, German, Spanish, and Chinese (Simplified and Traditional).
Supergraph enforces a complete, evidence-based coding pipeline — scan → plan → TDD → fix → verify → review — grounded in real codebase analysis at every step. It combines AST dependency graphs, LSP-level code intelligence, and a structured skill chain so Claude never guesses about impact before making a change.
Unity Development Toolkit - Expert agents for scripting/refactoring/optimization, script templates, and Agent Skills for Unity C# development