open-coleslaw 0.2.0 → 0.2.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (2) hide show
  1. package/README.md +87 -60
  2. package/package.json +1 -1
package/README.md CHANGED
@@ -2,46 +2,93 @@
2
2
 
3
3
  **Just type your prompt. Like coleslaw — it's already prepared, just scoop and eat.**
4
4
 
5
- Open Coleslaw is a multi-agent orchestrator plugin for Claude Code. It gives you an entire AI engineering team — architects, engineers, QA leads, product managers — that organizes itself, holds meetings, writes PRD-format minutes, and executes tasks. All you do is describe what you want.
5
+ Open Coleslaw is a multi-agent orchestrator plugin for Claude Code. It gives you an entire AI engineering team — architects, engineers, QA leads, product managers — that organizes itself, holds meetings, writes PRD-format minutes, and executes tasks.
6
6
 
7
7
  Zero commands to memorize. Zero tools to call manually. Your AI team is already hired.
8
8
 
9
9
  ## How It Works
10
10
 
11
- You type a prompt like *"Design an authentication system for our API"*. That's it. Here's what happens behind the scenes:
11
+ You type a prompt like *"Build me a balance game app"*. That's it.
12
12
 
13
- 1. **The Orchestrator** (your proxy) analyzes your request and decides which department leaders are needed
14
- 2. **Leaders convene a meeting** — Architecture, Engineering, QA discuss in structured rounds
15
- 3. **PRD meeting minutes** are automatically generated with decisions, action items, and technical specs
16
- 4. **Tasks are compacted** from minutes and assigned to departments
17
- 5. **Workers are hired** by each leader to execute tasks in parallel
18
- 6. **Results are reported** back through the hierarchy
13
+ ```
14
+ You: "Build me a balance game app"
15
+
16
+ → Orchestrator analyzes your request
17
+ → Convenes Architecture + Engineering + Product leaders
18
+ → Leaders hold a structured meeting (2-3 rounds per agenda item)
19
+ → PRD meeting minutes generated
20
+ → Plan presented to you for approval
21
+ → You: "OK"
22
+ → Tasks compacted → Workers hired → Code implemented
23
+ → Results reported
24
+ ```
19
25
 
20
- You get the final output. If agents need your input on an important decision, they'll `@mention` you.
26
+ You never call a tool. You never pick a department. You never manage an agent. The orchestrator handles everything. If agents need your input on an important decision, they `@mention` you.
21
27
 
22
28
  ## Installation
23
29
 
24
- ```bash
25
- # Add the marketplace
26
- /plugin marketplace add sshworld/open-coleslaw
30
+ In Claude Code:
27
31
 
28
- # Install the plugin
32
+ ```
33
+ /plugin marketplace add sshworld/open-coleslaw
29
34
  /plugin install open-coleslaw@sshworld
30
35
  ```
31
36
 
32
- ### Verify Installation
37
+ Then start a new session. That's it — every prompt now goes through the orchestrator pipeline.
38
+
39
+ ### Verify It Works
40
+
41
+ Start a new session and type anything:
42
+
43
+ ```
44
+ Design a REST API for a todo app
45
+ ```
46
+
47
+ You should see the orchestrator agent being dispatched and a meeting starting automatically.
48
+
49
+ ## The Pipeline
33
50
 
34
- Start a new Claude Code session and just type naturally:
51
+ Every request follows this flow. No exceptions.
35
52
 
36
53
  ```
37
- Design a REST API for a todo app with user authentication
54
+ Prompt → Orchestrator → Meeting → PRD Minutes → Your Approval → Implement
55
+ ```
56
+
57
+ 1. **Orchestrator dispatched** — analyzes your request, selects departments
58
+ 2. **Meeting convened** — leaders discuss in structured rounds
59
+ 3. **PRD minutes generated** — decisions, action items, technical specs
60
+ 4. **Plan presented** — you review and approve (or request changes)
61
+ 5. **Tasks compacted** — minutes converted to actionable tasks per department
62
+ 6. **Workers hired** — each leader deploys workers in parallel
63
+ 7. **Results reported** — final output delivered to you
64
+
65
+ If you want changes mid-plan, the orchestrator chains a follow-up meeting.
66
+
67
+ ## The Agent Hierarchy
68
+
69
+ ```
70
+ ┌─────────────────┐
71
+ │ Orchestrator │ ← Your proxy (NOT a CEO)
72
+ │ (claude-opus) │ Decides, but asks you
73
+ └────────┬────────┘ for important calls
74
+ ┌───────────┼───────────┐
75
+ ▼ ▼ ▼
76
+ ┌─────────┐ ┌─────────┐ ┌─────────┐
77
+ │ Arch │ │ Eng │ │ QA │ ← Leaders meet & discuss
78
+ │ Leader │ │ Leader │ │ Leader │ Then hire their own workers
79
+ └────┬────┘ └────┬────┘ └────┬────┘
80
+ ┌──┴──┐ ┌───┼───┐ ┌──┴──┐
81
+ │W1 W2│ │W3 W4 W5│ │W6 │ ← Workers execute in parallel
82
+ └─────┘ └───────┘ └─────┘
38
83
  ```
39
84
 
40
- If Open Coleslaw is working, the orchestrator will automatically convene a meeting with the right leaders and start the discussion.
85
+ **5 Departments**: Architecture, Engineering, QA, Product, Research
86
+
87
+ Each leader autonomously decides how many workers to hire based on task complexity.
41
88
 
42
- ## What You Get
89
+ ## What's Inside
43
90
 
44
- ### 14 MCP Tools (all automatic — you never call these directly)
91
+ ### 14 MCP Tools (orchestrator calls these — you never do)
45
92
 
46
93
  | Tool | What It Does |
47
94
  |------|-------------|
@@ -53,60 +100,39 @@ If Open Coleslaw is working, the orchestrator will automatically convene a meeti
53
100
  | `get-task-report` | Shows execution results per department |
54
101
  | `get-agent-tree` | Displays the full agent hierarchy |
55
102
  | `respond-to-mention` | Handles decisions the agents need from you |
56
- | `get-mentions` | Lists pending `@mention` decision points |
103
+ | `get-mentions` | Lists pending @mention decision points |
57
104
  | `cancel-meeting` | Stops a meeting and cascades to all workers |
58
105
  | `list-meetings` | Shows meeting history |
59
106
  | `create-capability` | Self-extends the plugin with new hooks/skills |
60
107
  | `get-cost-summary` | Tracks spending per agent, meeting, department |
61
108
  | `chain-meeting` | Links meetings — output of one feeds into the next |
62
109
 
63
- ### 6 Skills (AI decides when to use them)
110
+ ### 7 Skills
64
111
 
65
- | Skill | Triggers When |
66
- |-------|--------------|
67
- | `meeting` | You ask to design, plan, discuss, or review something |
68
- | `status` | You ask about ongoing work or what agents are doing |
69
- | `dashboard` | You want to see the real-time visualization |
70
- | `mention` | There are pending decisions that need your input |
71
- | `agents` | You ask about the team hierarchy or who's working on what |
72
- | `minutes` | You want to read past meeting minutes or search decisions |
112
+ | Skill | Purpose |
113
+ |-------|---------|
114
+ | `using-open-coleslaw` | Loaded at session start — ensures all requests go through the pipeline |
115
+ | `meeting` | Dispatches the orchestrator for the meeting → plan → implement flow |
116
+ | `status` | Shows active meetings, agents, pending mentions |
117
+ | `dashboard` | Opens the real-time Neon Ops Center |
118
+ | `mention` | Handles pending @mention decisions |
119
+ | `agents` | Shows the agent hierarchy tree |
120
+ | `minutes` | Browses past meeting minutes |
73
121
 
74
122
  ### Real-Time Dashboard
75
123
 
76
- A cyberpunk-themed "Neon Ops Center" runs at `http://localhost:35143`:
124
+ A cyberpunk-themed "Neon Ops Center" at `http://localhost:35143`:
77
125
 
78
126
  - Live agent hierarchy with animated connections
127
+ - Per-project tabs (multiple terminals → multiple tabs)
79
128
  - Meeting progress tracking
80
129
  - Task delegation and completion flow
81
- - `@mention` alerts
82
- - Cost tracking
130
+ - @mention alerts
131
+ - Duplicate project names get auto-numbered: `project`, `project (1)`
83
132
 
84
133
  ### Self-Extending
85
134
 
86
- If you ask for a workflow that doesn't exist yet, the orchestrator creates it — new hooks, skills, or automations — and registers them for future use.
87
-
88
- ## The Agent Hierarchy
89
-
90
- ```
91
- ┌─────────────────┐
92
- │ Orchestrator │ ← Your proxy (NOT a CEO)
93
- │ "What do you │ Decides, but asks you
94
- │ need built?" │ for important calls
95
- └────────┬────────┘
96
- ┌───────────┼───────────┐
97
- ▼ ▼ ▼
98
- ┌─────────┐ ┌─────────┐ ┌─────────┐
99
- │ Arch │ │ Eng │ │ QA │ ← Leaders meet & discuss
100
- │ Leader │ │ Leader │ │ Leader │ Then hire their own workers
101
- └────┬────┘ └────┬────┘ └────┬────┘
102
- ┌──┴──┐ ┌───┼───┐ ┌──┴──┐
103
- │W1 W2│ │W3 W4 W5│ │W6 │ ← Workers execute in parallel
104
- └─────┘ └───────┘ └─────┘
105
- ```
106
-
107
- **5 Departments**: Architecture, Engineering, QA, Product, Research
108
-
109
- Each leader autonomously decides how many workers to hire based on task complexity.
135
+ Ask for a workflow that doesn't exist yet, and the orchestrator creates it — new hooks, skills, or automations — registered for future use.
110
136
 
111
137
  ## Agent Tiers
112
138
 
@@ -125,10 +151,11 @@ Coleslaw is a side dish that's already made. You don't prepare it — you just e
125
151
 
126
152
  ### Key Decisions
127
153
 
128
- - **The Orchestrator is your proxy, not a CEO.** You are the decision-maker. The orchestrator acts on your behalf but escalates important choices via `@mention`.
154
+ - **The Orchestrator is your proxy, not a CEO.** You are the decision-maker. The orchestrator acts on your behalf but escalates important choices via @mention.
155
+ - **Meeting first, always.** Even "simple" requests go through the pipeline. If it's truly simple, the meeting will be fast.
129
156
  - **MVP cycles.** Work happens in loops: meeting → develop → verify → (re-meet if needed).
130
- - **Rules survive context compaction.** Core rules are injected every session via hooks, embedded in agent prompts, and stored in persistent files. The system never forgets how to behave.
131
- - **Agents check before they code.** Every agent analyzes the project's dependencies, existing code, and conventions before writing anything. No duplicate packages, no style violations.
157
+ - **Rules survive context compaction.** The `using-open-coleslaw` skill is injected at every session start. The system never forgets how to behave.
158
+ - **Agents check before they code.** Every agent analyzes the project's dependencies, existing code, and conventions before writing anything.
132
159
 
133
160
  ## Development
134
161
 
@@ -141,7 +168,7 @@ npm run build
141
168
  # Run with mock agents (no Claude CLI needed)
142
169
  COLESLAW_MOCK=1 node dist/index.js
143
170
 
144
- # Run tests
171
+ # Run tests (218 tests)
145
172
  npm test
146
173
 
147
174
  # Type check
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "open-coleslaw",
3
- "version": "0.2.0",
3
+ "version": "0.2.1",
4
4
  "description": "Multi-agent orchestrator MCP plugin for Claude Code with hierarchical agent system",
5
5
  "type": "module",
6
6
  "bin": {