popcoru

v0.7.1

Published

5 days ago

Agent Workspace OS — harness engineering for AI coding agents

Downloads

2,429

0High
0Medium
0Low

minkyungwook

ai coding-agent harness-engineering cli typescript lint verification

SUNCO

Agent Workspace OS — harness engineering for AI coding agents.

65 skills. 7-layer verification. Adaptive lifecycle. Zero mistakes.

npx popcoru

What is SUNCO?

In an era where AI agents write code, the builder's job is not writing code — it's setting up the field so agents make fewer mistakes. SUNCO is that field.

Harness Engineering — lint, health check, and guard enforce quality deterministically (zero LLM cost)
65 Skills — every capability is a skill. No hardcoded commands.
7-Layer Verification — multi-agent review, guardrails, BDD, permissions, adversarial, cross-model, human eval
Adaptive Lifecycle — pivot, rethink, backtrack, reinforce when requirements change
Multi-Runtime — Claude Code, Codex, Cursor, Antigravity

Install

npx popcoru

Interactive installer asks: language (EN/KR), runtime (Claude Code/Codex/Cursor/Antigravity/All).

Quick Start

/sunco:new           # idea -> questions -> research -> roadmap
/sunco:discuss 1     # extract decisions for Phase 1
/sunco:plan 1        # create atomic execution plans
/sunco:execute 1     # run plans in parallel waves
/sunco:verify 1      # 7-layer Swiss cheese verification
/sunco:ship 1        # create PR and release

Commands (65)

Harness (Zero LLM Cost)

| Command | Description | |---------|-------------| | init | Scaffold .sun/ and .planning/ | | lint | Architecture boundary check | | health | Codebase health score with trends | | guard | Real-time lint-on-change + rule drafting | | agents | Agent instruction analysis |

Workflow

| Command | Description | |---------|-------------| | new | Bootstrap from idea to roadmap | | discuss | Extract decisions before planning | | plan | Create verified execution plans | | execute | Wave-based parallel execution | | verify | 7-layer verification | | ship | Create PR after verification | | auto | Full autonomous pipeline |

Adaptive Lifecycle (NEW in v0.3.2)

| Command | Description | |---------|-------------| | pivot | Detect scope changes, re-route phases | | rethink | Revise specific decisions | | backtrack | Restore to rollback point | | reinforce | Add requirements mid-milestone | | where-am-i | Full orientation dashboard | | impact-analysis | Invalidation cascade | | design-pingpong | Cross-model merge + debate |

Debug

| Command | Description | |---------|-------------| | debug | Systematic debugging with state | | diagnose | Quick system diagnostic | | forensics | Post-mortem analysis |

Session

| Command | Description | |---------|-------------| | status | Current state | | progress | Phase completion % | | next | Auto-advance | | pause / resume | Session management | | where-am-i | Orientation dashboard |

Composition

| Command | Description | |---------|-------------| | auto | Full pipeline (discuss -> ship) | | quick | Fast execution with guarantees | | fast | Skip planning, execute immediately | | do | Natural language -> command routing | | mode | SUNCO super mode |

Architecture

sunco-tools.cjs (2,179 lines)    — Workflow automation engine
65 commands                        — Thin routers to workflows
64 workflows                       — Full process definitions
18 agents                          — Specialized AI agent definitions
15 references                      — Domain knowledge documents
25+ templates                      — Artifact templates
4 hooks                            — Git/lifecycle hooks

v0.3.2 Highlights

Adaptive Lifecycle — 7 new workflows for handling scope changes mid-project
Quality Supremacy — executor, phase-researcher, integration-checker agents enhanced
Artifact Hash System — detect changes to planning artifacts, compute invalidation cascade
Rollback Points — snapshot and restore .planning/ state at any phase boundary
Design Pingpong — cross-model merge + debate protocol (2.4x cost, opt-in)
Guard --draft-claude-rules — auto-generate conditional .claude/rules/ from codebase patterns
Codex Benchmark — A/B comparison framework for harness quality testing

Stats

| Category | Lines | |----------|-------| | Workflows | 25,306 | | Templates | 7,347 | | References | 5,026 | | Agents | 10,111 | | Tools | 2,529 | | Total | 50,319 |

License

MIT