@leclabs/agent-toolkit 2.1.0 → 3.0.0

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 +43 -88
  2. package/package.json +2 -2
package/README.md CHANGED
@@ -1,121 +1,76 @@
1
1
  # Agent Toolkit
2
2
 
3
- A Claude Code plugin marketplace for AI agent tools. The flagship plugin is **Flow** -- a stateless state-machine for Agents.
3
+ AI agent tools. Ships with **Navigator** (MCP server) and **Flow** (Claude Code plugin).
4
+
5
+ ## Components
6
+
7
+ | Component | Pattern | Provides |
8
+ | -------------------------------------------------------- | -------------------- | --------------------- |
9
+ | [Navigator](packages/agent-flow-navigator-mcp/README.md) | Finite State Machine | Workflow navigation |
10
+ | [Flow](plugins/flow/README.md) | Prompt injection | Orchestration ruleset |
11
+
12
+ **Inversion of control**: Traditional workflow engines use agents as tools. Navigator flips this - the AI orchestrator stays in control and uses the FSM for navigation.
4
13
 
5
14
  ## Installation
6
15
 
7
16
  ```bash
8
- # Add the toolkit to Claude Code
17
+ # Install Flow plugin (includes Navigator MCP)
9
18
  claude plugin marketplace add leclabs/agent-toolkit
10
-
11
19
  claude plugin install flow@agent-toolkit
12
20
  ```
13
21
 
14
- ```mermaid
15
- flowchart TD
16
- start(("Start"))
17
- work["Do Work<br/><small>Developer</small>"]
18
- check{"Check"}
19
- end_success[["Done"]]
20
- hitl_blocked{{"Blocked"}}
21
-
22
- start --> work
23
- work --> check
24
- check -->|passed| end_success
25
- check -->|failed| work
26
- check -->|failed| hitl_blocked
27
- hitl_blocked -->|passed| work
28
-
29
- classDef startStep fill:#90EE90,stroke:#228B22
30
- classDef successStep fill:#87CEEB,stroke:#4169E1
31
- classDef hitlStep fill:#FFB6C1,stroke:#DC143C
32
- classDef gateStep fill:#E6E6FA,stroke:#9370DB
33
- classDef currentStep fill:#FFD700,stroke:#FF8C00,stroke-width:3px
34
- class start startStep
35
- class end_success successStep
36
- class hitl_blocked hitlStep
37
- class check gateStep
38
- ```
39
-
40
- ## Quick Start
22
+ ## Quick Start (Flow)
41
23
 
42
24
  ```bash
43
- # Create a task with a workflow
44
25
  /flow:task "add user authentication"
45
-
46
- # Execute all pending tasks
47
26
  /flow:go
48
27
  ```
49
28
 
50
- ## Commands
51
-
52
- | Command | Description |
53
- | ------------- | -------------------------------------------- |
54
- | `/flow:task` | Create a task and choose a workflow |
55
- | `/flow:go` | Execute all pending tasks |
56
- | `/flow:recon` | Deep project reconnaissance |
57
- | `/flow:list` | List available workflows |
58
- | `/flow:setup` | Set up workflows and agents for your project |
59
-
60
- ## Workflows
61
-
62
- 14 workflow templates ship in the catalog:
63
-
64
- | Workflow | Steps | Description |
65
- | ------------------------- | ----- | ------------------------------------------------------------------ |
66
- | feature-development | 15 | Full lifecycle: plan, implement, test, review, PR |
67
- | bug-fix | 11 | Reproduce, investigate, fix, regression test |
68
- | bug-hunt | 16 | Parallel investigation: reproduce, code archaeology, git forensics |
69
- | agile-task | 9 | General task: analyze, implement, test, review |
70
- | quick-task | 8 | Minimal: understand, execute, verify |
71
- | test-coverage | 10 | Analyze gaps, write tests, review |
72
- | context-gather | 10 | Parallel context gathering: repo, system, weather |
73
- | context-optimization | 10 | Map connections, identify pathologies, improve |
74
- | ui-reconstruction | 17 | Extract semantic IR, rebuild UI, blind review |
75
- | refactor | 16 | Functional core / imperative shell restructuring |
76
- | build-review-murder-board | 7 | Build-review loop, level 5 scrutiny, blind-shot review |
77
- | build-review-quick | 7 | Build-review loop, basic sanity check |
78
- | execute | 3 | Single-step workflow: just do the thing |
79
- | hitl-test | 5 | Minimal HITL recovery test: work, gate, escalate |
80
-
81
- Customize workflows for your project with `/flow:setup`.
82
-
83
29
  ## Architecture
84
30
 
85
31
  ```mermaid
86
32
  flowchart TB
87
- user["/flow:task 'add dark mode'"]
88
-
89
- subgraph Orchestrator
90
- start["flow:start"]
33
+ subgraph "Claude Code Plugin"
34
+ orchestrator["Orchestration Layer"]
91
35
  end
92
36
 
93
- subgraph Navigator["Navigator (MCP)"]
37
+ subgraph Navigator["Navigator (MCP Server)"]
94
38
  sm["State Machine"]
39
+ catalog["Workflow Catalog"]
95
40
  end
96
41
 
97
- subgraph Subagents
98
- planner["Planner"]
99
- developer["Developer"]
100
- tester["Tester"]
101
- reviewer["Reviewer"]
102
- end
103
-
104
- subgraph Workflows
105
- wf["Graph Definitions"]
42
+ subgraph "Your Project"
43
+ workflows["Custom Workflows"]
44
+ agents["Agent Definitions"]
106
45
  end
107
46
 
108
- user --> Orchestrator
109
- Orchestrator <-->|MCP| Navigator
110
- Orchestrator --> Subagents
111
- Navigator --> Workflows
47
+ orchestrator <-->|MCP Protocol| Navigator
48
+ Navigator --> catalog
49
+ Navigator --> workflows
112
50
  ```
113
51
 
114
- ## Links
52
+ ## Workflows
115
53
 
116
- - [Flow Plugin](plugins/flow/README.md) -- commands, skills, workflows, and customization
117
- - [Navigator MCP Server](packages/agent-flow-navigator-mcp/README.md) -- workflow state machine
118
- - [Workflow Diagrams](.flow/diagrams/) -- mermaid diagrams for all workflows
54
+ 16 workflow templates in the catalog:
55
+
56
+ | Workflow | Purpose |
57
+ | ------------------------- | --------------------------------------------------------- |
58
+ | feature-development | Full lifecycle: plan, implement, test, review, PR |
59
+ | bug-fix | Reproduce, investigate, fix, regression test |
60
+ | bug-hunt | Parallel investigation: reproduce, archaeology, forensics |
61
+ | quick-task | Minimal: understand, execute, verify |
62
+ | agile-task | General: analyze, implement, test, review |
63
+ | test-coverage | Analyze gaps, write tests, review |
64
+ | refactor | Functional core / imperative shell restructuring |
65
+ | ui-reconstruction | Extract semantic IR, rebuild UI, blind review |
66
+ | context-optimization | Map connections, identify pathologies, improve |
67
+ | build-review-murder-board | Build-review loop with 80% approval threshold |
68
+ | build-review-quick | Build-review loop with basic sanity check |
69
+ | batch-process | Dynamic fork/join for processing multiple items |
70
+ | context-gather | Parallel context gathering |
71
+ | execute | Single-step: just do the thing |
72
+ | odd-even-test | Test workflow demonstrating fork with retry logic |
73
+ | hitl-test | Minimal HITL recovery test |
119
74
 
120
75
  ## License
121
76
 
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@leclabs/agent-toolkit",
3
- "version": "2.1.0",
3
+ "version": "3.0.0",
4
4
  "description": "The essential Claude Code plugin marketplace for reliable Agent-led workflow orchestration",
5
5
  "main": "index.js",
6
6
  "bin": {
@@ -17,7 +17,7 @@
17
17
  "version": "changeset version && node scripts/sync-versions.js",
18
18
  "release": "npm whoami --registry https://registry.npmjs.org/ || (echo '\\nAuth failed. Run: npm config set //registry.npmjs.org/:_authToken=YOUR_TOKEN' && exit 1) && npm publish --access public && cd packages/agent-flow-navigator-mcp && npm publish --access public",
19
19
  "reload": "npm run reload:mcp && npm run reload:plugin",
20
- "reload:mcp": "npm link --prefix packages/agent-flow-navigator-mcp && npm link @leclabs/agent-flow-navigator-mcp",
20
+ "reload:mcp": "(cd packages/agent-flow-navigator-mcp && npm link) && npm link @leclabs/agent-flow-navigator-mcp",
21
21
  "reload:plugin": "(claude plugin uninstall flow@agent-toolkit || true) && (claude plugin marketplace remove agent-toolkit || true) && claude plugin marketplace add $(pwd) && claude plugin install flow@agent-toolkit"
22
22
  },
23
23
  "repository": {