Skip to content

Instantly share code, notes, and snippets.

@jleechan2015
Created March 21, 2026 02:06
Show Gist options
  • Select an option

  • Save jleechan2015/4bbb69bf5d313e0a4677e3e2f3a03028 to your computer and use it in GitHub Desktop.

Select an option

Save jleechan2015/4bbb69bf5d313e0a4677e3e2f3a03028 to your computer and use it in GitHub Desktop.
jleechanclaw + agento presentation (v10) — for Gamma

jleechanclaw + agento

From idea to merged PR — autonomous AI-powered development


The Problem

  • AI coding tools are powerful, but operationally noisy
  • Every task requires manual handoffs: chat → code → CI → review → fix → repeat
  • Engineers spend more time shepherding PRs than writing code
  • Goal: "Idea in, merged PR out" with minimal manual overhead

How It Works

  1. You describe what you want in Slack (natural language)
  2. jleechanclaw expands your intent with repo context, memory, and policy
  3. agento spawns isolated worktrees and delegates to coding agents
  4. CI, code review, and fix loops run automatically until stable
  5. You get a proof-first status update: PR URL + commits + green checks

System Roles

jleechanclaw (OpenClaw Harness)

Understands intent, expands context, routes work, manages memory, sends status updates

agento (Agent Orchestrator)

Spawns isolated coding sessions, runs test/fix loops, iterates PRs to green

GitHub + CI + Review Bots

CodeRabbit, Bugbot, and Copilot provide objective quality signals and automated merge gates


Design Principles

  • Config over code — behavior changes via YAML and policy files, not new Python
  • Evidence-first — every status update includes PR URLs, commit SHAs, and check results
  • Fail-closed gates — 6-point merge criteria must all pass before auto-merge
  • Isolated execution — each task runs in its own git worktree, no cross-contamination
  • Human-in-the-loop — agents propose, humans approve merges

Repo-Backed Configuration

All config is git-tracked in jleechanorg/jleechanclaw:

  • Routing & dispatch rules — agent-orchestrator.yaml
  • Agent behavior & constraints — SOUL.md, TOOLS.md, CLAUDE.md
  • Health monitoring — monitor-agent.sh (launchd loop)
  • Notification pipeline — src/orchestration/openclaw_notifier.py
  • Design docs — roadmap/NATURAL_LANGUAGE_DISPATCH.md

System Diagram

┌─────────────────────┐     ┌──────────────────────────────┐
│   User (Slack)       │     │ jleechanclaw (OpenClaw)       │
│                      │────▶│                              │
│ Natural-language     │     │ Context expansion            │
│ request + decisions  │     │ Memory + policy + routing    │
└─────────────────────┘     │ Evidence-first status updates│
                            └──────────┬───────────────────┘
                                       │
                                       ▼
┌─────────────────────┐     ┌──────────────────────────────┐     ┌─────────────────────┐
│ agento / Agent-      │     │ GitHub + CI + Review Bots    │     │     Outcome          │
│ Orchestrator         │────▶│                              │────▶│                     │
│                      │     │ PR checks, code review,      │     │ Merge-ready PR      │
│ Spawns worktrees     │     │ mergeability + quality       │     │ (URL + commits +    │
│ Delegates agents     │     │ signals                      │     │  checks)            │
│ Automates CI/review  │     │                              │     │ Human final merge   │
└─────────────────────┘     └──────────────────────────────┘     └─────────────────────┘

Get In Touch

  • GitHub: github.com/jleechanorg
  • LinkedIn: linkedin.com/in/jeffrey-lee-chan
  • Repo: github.com/jleechanorg/jleechanclaw
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment