planflow-ai 1.0.2 → 1.0.3
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.
- package/.claude/commands/create-plan.md +29 -11
- package/.claude/commands/discovery-plan.md +25 -12
- package/.claude/commands/execute-plan.md +34 -12
- package/.claude/commands/flow.md +105 -0
- package/.claude/commands/review-code.md +7 -1
- package/.claude/rules/core/autopilot-mode.md +188 -0
- package/.claude/rules/patterns/plans-patterns.md +2 -2
- package/README.md +34 -1
- package/package.json +1 -1
- package/rules/core/_index.mdc +14 -2
- package/rules/core/autopilot-mode.mdc +193 -0
- package/rules/patterns/plans-patterns.mdc +2 -2
- package/rules/skills/create-plan-skill.mdc +10 -13
- package/skills/plan-flow/SKILL.md +2 -1
- package/skills/plan-flow/flow/SKILL.md +81 -0
- package/templates/shared/AGENTS.md.template +2 -1
- package/templates/shared/CLAUDE.md.template +3 -2
|
@@ -58,12 +58,19 @@ RELATED COMMANDS:
|
|
|
58
58
|
|
|
59
59
|
---
|
|
60
60
|
|
|
61
|
+
> ⚠️ **AUTOPILOT MODE CHECK**
|
|
62
|
+
>
|
|
63
|
+
> Before proceeding, check if `flow/.autopilot` exists.
|
|
64
|
+
> - **If YES**: Autopilot is ON. After creating the plan and getting user approval, **auto-proceed** to `/execute-plan` with the plan output. Do NOT wait for manual invocation.
|
|
65
|
+
> - **If NO**: Follow the standard rules below (stop and wait for user).
|
|
66
|
+
|
|
61
67
|
## Critical Rules
|
|
62
68
|
|
|
63
69
|
| Rule | Description |
|
|
64
70
|
| ------------------------ | -------------------------------------------------------- |
|
|
65
|
-
| **
|
|
66
|
-
| **
|
|
71
|
+
| **Discovery Required** | NEVER create a plan without a discovery document. If none exists, run `/discovery-plan` first. No exceptions. |
|
|
72
|
+
| **No Auto-Chaining** | NEVER auto-invoke /execute-plan - user must invoke it (unless autopilot ON) |
|
|
73
|
+
| **Complete and Stop** | After presenting results, STOP and wait for user (unless autopilot ON) |
|
|
67
74
|
|
|
68
75
|
---
|
|
69
76
|
|
|
@@ -80,18 +87,27 @@ RELATED COMMANDS:
|
|
|
80
87
|
|
|
81
88
|
---
|
|
82
89
|
|
|
83
|
-
### Step 2: Validate Discovery Phase Completion
|
|
90
|
+
### Step 2: Validate Discovery Phase Completion (HARD BLOCK)
|
|
91
|
+
|
|
92
|
+
**A discovery document is REQUIRED before creating any plan. No exceptions.**
|
|
84
93
|
|
|
85
|
-
|
|
94
|
+
1. Check user input for a discovery document reference (`@flow/discovery/...`)
|
|
95
|
+
2. If no reference provided, search `flow/discovery/` for a matching discovery document
|
|
96
|
+
3. **If NO discovery document exists**: **STOP immediately.** Do NOT create a plan. Instead:
|
|
97
|
+
- Inform the user that discovery is required before planning
|
|
98
|
+
- Invoke `/discovery-plan` to start the discovery process
|
|
99
|
+
- Example response:
|
|
100
|
+
```
|
|
101
|
+
A discovery document is required before creating a plan. Discovery refines your idea,
|
|
102
|
+
gathers requirements, and identifies risks — skipping it leads to incomplete plans.
|
|
86
103
|
|
|
87
|
-
|
|
88
|
-
|
|
89
|
-
-
|
|
90
|
-
|
|
91
|
-
3. If NO discovery indicators found: Recommend `/discovery-plan` command first
|
|
92
|
-
4. If discovery indicators ARE found: Proceed with plan creation
|
|
104
|
+
Starting discovery now...
|
|
105
|
+
```
|
|
106
|
+
- Then run the discovery process for the user's feature
|
|
107
|
+
4. If a discovery document IS found: Proceed with plan creation using that document
|
|
93
108
|
|
|
94
109
|
**Important**: NEVER read or reference files in `flow/archive/` - these are outdated.
|
|
110
|
+
**Important**: NEVER create a plan without a discovery document. This rule has NO exceptions.
|
|
95
111
|
|
|
96
112
|
---
|
|
97
113
|
|
|
@@ -138,7 +154,9 @@ Plan Created!
|
|
|
138
154
|
3. When ready, invoke `/execute-plan @flow/plans/plan_<feature>_v<version>.md`
|
|
139
155
|
```
|
|
140
156
|
|
|
141
|
-
**CRITICAL**:
|
|
157
|
+
**CRITICAL**: If autopilot mode is OFF (`flow/.autopilot` does not exist), this command is now complete. Do NOT auto-invoke `/execute-plan`. Wait for the user to explicitly invoke it.
|
|
158
|
+
|
|
159
|
+
**If autopilot mode is ON** (`flow/.autopilot` exists): Auto-proceed to `/execute-plan` with the plan file after the user approves the plan.
|
|
142
160
|
|
|
143
161
|
---
|
|
144
162
|
|
|
@@ -12,15 +12,21 @@ This command creates a discovery document for gathering and clarifying requireme
|
|
|
12
12
|
|
|
13
13
|
---
|
|
14
14
|
|
|
15
|
+
> ⚠️ **AUTOPILOT MODE CHECK**
|
|
16
|
+
>
|
|
17
|
+
> Before proceeding, check if `flow/.autopilot` exists.
|
|
18
|
+
> - **If YES**: Autopilot is ON. After completing discovery and user Q&A, **auto-proceed** to `/create-plan` with the discovery output. Do NOT ask "Would you like to proceed?" - just continue.
|
|
19
|
+
> - **If NO**: Follow the standard rules below (stop and wait for user).
|
|
20
|
+
|
|
15
21
|
> ⚠️ **IMPORTANT - DISCOVERY ONLY**
|
|
16
22
|
>
|
|
17
23
|
> This command ONLY creates a discovery document. It does NOT:
|
|
18
24
|
> - Create implementation plans (use `/create-plan` for that)
|
|
19
25
|
> - Execute code or implementation steps (use `/execute-plan` for that)
|
|
20
26
|
> - Write any source code files
|
|
21
|
-
> - Auto-chain to the next command
|
|
27
|
+
> - Auto-chain to the next command (unless autopilot mode is ON)
|
|
22
28
|
>
|
|
23
|
-
> After creating the discovery document, the command STOPS and waits for user review.
|
|
29
|
+
> After creating the discovery document, the command STOPS and waits for user review (unless autopilot mode is ON).
|
|
24
30
|
|
|
25
31
|
---
|
|
26
32
|
|
|
@@ -85,8 +91,8 @@ These rules are mandatory. For detailed patterns and guidelines, see `.claude/ru
|
|
|
85
91
|
| **Read First** | Read all referenced documents before asking questions |
|
|
86
92
|
| **High-Level Only** | Technical considerations are conceptual, not code |
|
|
87
93
|
| **Allow Refinement** | User reviews and refines discovery before plan creation |
|
|
88
|
-
| **Ask Before Proceeding**| ASK user if they want to proceed to /create-plan
|
|
89
|
-
| **No Auto-Execution** | Do NOT auto-invoke commands without user confirmation
|
|
94
|
+
| **Ask Before Proceeding**| ASK user if they want to proceed to /create-plan (unless autopilot ON) |
|
|
95
|
+
| **No Auto-Execution** | Do NOT auto-invoke commands without user confirmation (unless autopilot ON) |
|
|
90
96
|
| **NO PLANNING** | Do NOT create implementation plans during discovery |
|
|
91
97
|
| **NO EXECUTION** | Do NOT execute any implementation steps |
|
|
92
98
|
|
|
@@ -177,14 +183,21 @@ Discovery Complete!
|
|
|
177
183
|
3. Proceed to planning when ready
|
|
178
184
|
```
|
|
179
185
|
|
|
180
|
-
**Now ask the user**:
|
|
181
|
-
|
|
182
|
-
```
|
|
183
|
-
|
|
184
|
-
|
|
185
|
-
|
|
186
|
-
|
|
187
|
-
|
|
186
|
+
**Now ask the user using the `AskUserQuestion` tool**:
|
|
187
|
+
|
|
188
|
+
```typescript
|
|
189
|
+
AskUserQuestion({
|
|
190
|
+
questions: [{
|
|
191
|
+
question: "Would you like me to proceed with creating the implementation plan?",
|
|
192
|
+
header: "Next step",
|
|
193
|
+
options: [
|
|
194
|
+
{ label: "Yes, create plan", description: "I'll invoke /create-plan with the discovery document" },
|
|
195
|
+
{ label: "No, review first", description: "You can review the discovery and invoke /create-plan when ready" },
|
|
196
|
+
{ label: "Refine discovery", description: "Let me know what needs to be adjusted in the discovery" }
|
|
197
|
+
],
|
|
198
|
+
multiSelect: false
|
|
199
|
+
}]
|
|
200
|
+
})
|
|
188
201
|
```
|
|
189
202
|
|
|
190
203
|
⚠️ **WAIT for user response before proceeding**
|
|
@@ -73,13 +73,19 @@ RELATED COMMANDS:
|
|
|
73
73
|
|
|
74
74
|
---
|
|
75
75
|
|
|
76
|
+
> ⚠️ **AUTOPILOT MODE CHECK**
|
|
77
|
+
>
|
|
78
|
+
> Before proceeding, check if `flow/.autopilot` exists.
|
|
79
|
+
> - **If YES**: Autopilot is ON. After completing execution (build + test pass), **auto-proceed** to `/review-code`. Do NOT stop and wait.
|
|
80
|
+
> - **If NO**: Follow the standard rules below (stop and wait for user).
|
|
81
|
+
|
|
76
82
|
## Critical Rules
|
|
77
83
|
|
|
78
84
|
| Rule | Description |
|
|
79
85
|
| ------------------------ | -------------------------------------------------------- |
|
|
80
86
|
| **Build ONLY at End** | Do NOT run build after each phase - only at the very end |
|
|
81
87
|
| **No Direct DB/ORM** | NEVER run ORM or database commands directly |
|
|
82
|
-
| **Complete and Stop** | After execution, STOP and wait for user
|
|
88
|
+
| **Complete and Stop** | After execution, STOP and wait for user (unless autopilot ON) |
|
|
83
89
|
|
|
84
90
|
---
|
|
85
91
|
|
|
@@ -182,7 +188,7 @@ After the skill completes:
|
|
|
182
188
|
|
|
183
189
|
1. Present summary of all changes made
|
|
184
190
|
2. Confirm all tests pass
|
|
185
|
-
3. Ask if the plan should be archived:
|
|
191
|
+
3. Ask if the plan should be archived using the `AskUserQuestion` tool:
|
|
186
192
|
|
|
187
193
|
```markdown
|
|
188
194
|
Execution Complete!
|
|
@@ -191,14 +197,20 @@ Execution Complete!
|
|
|
191
197
|
- X phases completed
|
|
192
198
|
- All tests passing
|
|
193
199
|
- Build successful
|
|
194
|
-
|
|
195
|
-
**Archive the plan?**
|
|
196
|
-
This will move the plan to `flow/archive/`:
|
|
197
|
-
```bash
|
|
198
|
-
mv flow/plans/plan_<feature>_v1.md flow/archive/
|
|
199
200
|
```
|
|
200
201
|
|
|
201
|
-
|
|
202
|
+
```typescript
|
|
203
|
+
AskUserQuestion({
|
|
204
|
+
questions: [{
|
|
205
|
+
question: "Would you like to archive this completed plan?",
|
|
206
|
+
header: "Archive",
|
|
207
|
+
options: [
|
|
208
|
+
{ label: "Yes, archive it", description: "Move the plan to flow/archive/ - recommended for completed plans" },
|
|
209
|
+
{ label: "No, keep it", description: "Keep the plan in flow/plans/ for reference" }
|
|
210
|
+
],
|
|
211
|
+
multiSelect: false
|
|
212
|
+
}]
|
|
213
|
+
})
|
|
202
214
|
```
|
|
203
215
|
|
|
204
216
|
---
|
|
@@ -566,10 +578,20 @@ npm run build && npm run test
|
|
|
566
578
|
|
|
567
579
|
4. **List all key changes** made
|
|
568
580
|
|
|
569
|
-
5. **Ask if the plan should be archived
|
|
570
|
-
|
|
571
|
-
```
|
|
572
|
-
|
|
581
|
+
5. **Ask if the plan should be archived** using the `AskUserQuestion` tool:
|
|
582
|
+
|
|
583
|
+
```typescript
|
|
584
|
+
AskUserQuestion({
|
|
585
|
+
questions: [{
|
|
586
|
+
question: "Would you like to archive this completed plan?",
|
|
587
|
+
header: "Archive",
|
|
588
|
+
options: [
|
|
589
|
+
{ label: "Yes, archive it", description: "Move the plan to flow/archive/" },
|
|
590
|
+
{ label: "No, keep it", description: "Keep the plan in flow/plans/" }
|
|
591
|
+
],
|
|
592
|
+
multiSelect: false
|
|
593
|
+
}]
|
|
594
|
+
})
|
|
573
595
|
```
|
|
574
596
|
|
|
575
597
|
---
|
|
@@ -0,0 +1,105 @@
|
|
|
1
|
+
---
|
|
2
|
+
description: Toggle autopilot flow mode - automatically orchestrates discovery, planning, execution, and review for feature requests
|
|
3
|
+
---
|
|
4
|
+
|
|
5
|
+
# Flow: Autopilot Mode Toggle
|
|
6
|
+
|
|
7
|
+
## Command Description
|
|
8
|
+
|
|
9
|
+
This command enables or disables **autopilot flow mode**. When enabled, every new user input is automatically classified and, if it's a feature request, the full plan-flow workflow runs automatically:
|
|
10
|
+
|
|
11
|
+
**contracts check → discovery → plan → execute → review-code → archive**
|
|
12
|
+
|
|
13
|
+
The LLM pauses only at mandatory checkpoints (discovery Q&A, plan approval).
|
|
14
|
+
|
|
15
|
+
---
|
|
16
|
+
|
|
17
|
+
## Help
|
|
18
|
+
|
|
19
|
+
**If the user invokes this command with `-help`, display only this section and stop:**
|
|
20
|
+
|
|
21
|
+
```
|
|
22
|
+
/flow - Autopilot Flow Mode Toggle
|
|
23
|
+
|
|
24
|
+
DESCRIPTION:
|
|
25
|
+
Enables or disables autopilot flow mode. When enabled, feature requests
|
|
26
|
+
automatically run the full plan-flow workflow without manual command invocation.
|
|
27
|
+
|
|
28
|
+
USAGE:
|
|
29
|
+
/flow -enable Enable autopilot mode (persists across sessions)
|
|
30
|
+
/flow -disable Disable autopilot mode
|
|
31
|
+
/flow -status Show current autopilot state
|
|
32
|
+
/flow Same as -status
|
|
33
|
+
/flow -help Show this help
|
|
34
|
+
|
|
35
|
+
BEHAVIOR WHEN ENABLED:
|
|
36
|
+
- Feature requests → full flow (discovery → plan → execute → review → archive)
|
|
37
|
+
- Trivial tasks (complexity 0-2) → executed directly, no flow
|
|
38
|
+
- Questions/exploration → answered normally
|
|
39
|
+
- Slash commands → run as normal
|
|
40
|
+
|
|
41
|
+
MANDATORY CHECKPOINTS (even in autopilot):
|
|
42
|
+
- Discovery phase: pauses for user Q&A
|
|
43
|
+
- Plan review: pauses for user approval before execution
|
|
44
|
+
|
|
45
|
+
CONTEXT MANAGEMENT:
|
|
46
|
+
- Only loads the relevant command context at each step
|
|
47
|
+
- Suggests context cleanup (/clear) after each completed flow cycle
|
|
48
|
+
|
|
49
|
+
STATE:
|
|
50
|
+
Persisted in flow/.autopilot (survives session restarts)
|
|
51
|
+
```
|
|
52
|
+
|
|
53
|
+
---
|
|
54
|
+
|
|
55
|
+
## Instructions
|
|
56
|
+
|
|
57
|
+
### When invoked with `-enable`
|
|
58
|
+
|
|
59
|
+
1. Create the file `flow/.autopilot` (empty file)
|
|
60
|
+
2. Confirm to the user:
|
|
61
|
+
|
|
62
|
+
```markdown
|
|
63
|
+
Autopilot flow mode **enabled**.
|
|
64
|
+
|
|
65
|
+
From now on, feature requests will automatically run the full workflow:
|
|
66
|
+
1. Check contracts → 2. Discovery (pause for Q&A) → 3. Create plan (pause for approval) → 4. Execute plan → 5. Review code → 6. Archive
|
|
67
|
+
|
|
68
|
+
Trivial tasks and questions are handled normally without the flow.
|
|
69
|
+
|
|
70
|
+
Use `/flow -disable` to turn off, or `/flow -status` to check state.
|
|
71
|
+
```
|
|
72
|
+
|
|
73
|
+
### When invoked with `-disable`
|
|
74
|
+
|
|
75
|
+
1. Delete the file `flow/.autopilot` (if it exists)
|
|
76
|
+
2. Confirm to the user:
|
|
77
|
+
|
|
78
|
+
```markdown
|
|
79
|
+
Autopilot flow mode **disabled**.
|
|
80
|
+
|
|
81
|
+
Commands will no longer auto-chain. Use individual slash commands as before:
|
|
82
|
+
`/discovery-plan` → `/create-plan` → `/execute-plan` → `/review-code`
|
|
83
|
+
```
|
|
84
|
+
|
|
85
|
+
### When invoked with `-status` or no arguments
|
|
86
|
+
|
|
87
|
+
1. Check if `flow/.autopilot` exists
|
|
88
|
+
2. Report:
|
|
89
|
+
|
|
90
|
+
```markdown
|
|
91
|
+
Autopilot flow mode: **[ENABLED/DISABLED]**
|
|
92
|
+
|
|
93
|
+
[If enabled]: Feature requests will automatically run the full workflow.
|
|
94
|
+
[If disabled]: Use individual slash commands to run each step manually.
|
|
95
|
+
```
|
|
96
|
+
|
|
97
|
+
---
|
|
98
|
+
|
|
99
|
+
## Critical Rules
|
|
100
|
+
|
|
101
|
+
| Rule | Description |
|
|
102
|
+
| --- | --- |
|
|
103
|
+
| **File-based state** | State is stored in `flow/.autopilot` - create to enable, delete to disable |
|
|
104
|
+
| **No other side effects** | This command ONLY manages the marker file. It does not run any workflow steps. |
|
|
105
|
+
| **Complete and stop** | After toggling state, STOP and wait for user input |
|
|
@@ -62,12 +62,18 @@ RELATED COMMANDS:
|
|
|
62
62
|
|
|
63
63
|
---
|
|
64
64
|
|
|
65
|
+
> ⚠️ **AUTOPILOT MODE CHECK**
|
|
66
|
+
>
|
|
67
|
+
> Before proceeding, check if `flow/.autopilot` exists.
|
|
68
|
+
> - **If YES**: Autopilot is ON. After completing the review, **auto-archive** the discovery and plan documents to `flow/archive/`, present the completion summary, and prompt for context cleanup (`/clear`).
|
|
69
|
+
> - **If NO**: Follow the standard rules below (stop and wait for user).
|
|
70
|
+
|
|
65
71
|
## Critical Rules
|
|
66
72
|
|
|
67
73
|
| Rule | Description |
|
|
68
74
|
| ------------------------ | -------------------------------------------------------- |
|
|
69
75
|
| **Read-Only Analysis** | This command ONLY produces a review document |
|
|
70
|
-
| **Complete and Stop** | After presenting results, STOP and wait for user
|
|
76
|
+
| **Complete and Stop** | After presenting results, STOP and wait for user (unless autopilot ON) |
|
|
71
77
|
|
|
72
78
|
---
|
|
73
79
|
|
|
@@ -0,0 +1,188 @@
|
|
|
1
|
+
# Autopilot Flow Mode
|
|
2
|
+
|
|
3
|
+
## Purpose
|
|
4
|
+
|
|
5
|
+
When `flow/.autopilot` exists, this rule is active. It automatically orchestrates the full plan-flow workflow for feature requests, loading only the relevant command context at each step.
|
|
6
|
+
|
|
7
|
+
---
|
|
8
|
+
|
|
9
|
+
## On Session Start
|
|
10
|
+
|
|
11
|
+
1. Check if `flow/.autopilot` exists
|
|
12
|
+
2. If it exists: autopilot mode is **ON** - follow the rules below for every user input
|
|
13
|
+
3. If it does not exist: autopilot mode is **OFF** - ignore this rule entirely, normal behavior applies
|
|
14
|
+
|
|
15
|
+
---
|
|
16
|
+
|
|
17
|
+
## Input Classification
|
|
18
|
+
|
|
19
|
+
When autopilot is ON, classify every new user input before acting:
|
|
20
|
+
|
|
21
|
+
### Category 1: Slash Command
|
|
22
|
+
**Signals**: Input starts with `/` (e.g., `/review-pr`, `/write-tests`, `/flow -disable`)
|
|
23
|
+
**Action**: Run that command normally. Do NOT trigger the flow.
|
|
24
|
+
|
|
25
|
+
### Category 2: Question or Exploration
|
|
26
|
+
**Signals**: User is asking about code, requesting explanations, exploring the codebase, or asking for help.
|
|
27
|
+
**Examples**: "What does this function do?", "How does auth work here?", "Show me the API routes"
|
|
28
|
+
**Action**: Answer normally. Do NOT trigger the flow.
|
|
29
|
+
|
|
30
|
+
### Category 3: Trivial Task
|
|
31
|
+
**Signals**: Single-file change, obvious fix, user specifies exact change, no architectural decisions needed. Estimated complexity 0-2.
|
|
32
|
+
**Examples**: "Fix the typo in README", "Rename this variable to camelCase", "Add a missing import"
|
|
33
|
+
**Action**: Execute directly. Do NOT trigger the flow.
|
|
34
|
+
|
|
35
|
+
### Category 4: Feature Request
|
|
36
|
+
**Signals**: Adding new functionality, behavior changes, multi-file scope, mentions creating/building/implementing, references APIs/integrations/components. Estimated complexity 3+.
|
|
37
|
+
**Examples**: "Add dark mode support", "Implement user authentication", "Create a new API endpoint for payments"
|
|
38
|
+
**Action**: **Trigger the full flow** (see below).
|
|
39
|
+
|
|
40
|
+
### When Uncertain
|
|
41
|
+
If the input doesn't clearly fit a category, use the `AskUserQuestion` tool:
|
|
42
|
+
|
|
43
|
+
```typescript
|
|
44
|
+
AskUserQuestion({
|
|
45
|
+
questions: [{
|
|
46
|
+
question: "I'm in autopilot mode. How should I handle this request?",
|
|
47
|
+
header: "Flow mode",
|
|
48
|
+
options: [
|
|
49
|
+
{ label: "Run full flow", description: "Run the full workflow: discovery → plan → execute → review" },
|
|
50
|
+
{ label: "Handle directly", description: "Just handle this task directly without the full flow" }
|
|
51
|
+
],
|
|
52
|
+
multiSelect: false
|
|
53
|
+
}]
|
|
54
|
+
})
|
|
55
|
+
```
|
|
56
|
+
|
|
57
|
+
---
|
|
58
|
+
|
|
59
|
+
## Flow Execution Steps
|
|
60
|
+
|
|
61
|
+
When a **feature request** is detected, execute these steps in order. At each step, **Read the relevant command file** to load its context, then follow its instructions.
|
|
62
|
+
|
|
63
|
+
### Step 1: Check Contracts
|
|
64
|
+
|
|
65
|
+
Check `flow/contracts/` for any relevant integration contracts. If found, note them for the discovery step.
|
|
66
|
+
|
|
67
|
+
### Step 2: Discovery
|
|
68
|
+
|
|
69
|
+
1. **Read** the file `.claude/commands/discovery-plan.md` to load discovery context
|
|
70
|
+
2. Execute the discovery skill following that command's instructions
|
|
71
|
+
3. **PAUSE**: Ask user discovery questions as needed (mandatory checkpoint)
|
|
72
|
+
4. Produce the discovery document in `flow/discovery/`
|
|
73
|
+
5. Write transition summary (see format below)
|
|
74
|
+
6. **Auto-proceed** to Step 3 (override the "ask before proceeding" rule)
|
|
75
|
+
|
|
76
|
+
### Step 3: Create Plan
|
|
77
|
+
|
|
78
|
+
**GATE CHECK**: Before proceeding, verify that Step 2 produced a discovery document in `flow/discovery/`. If no discovery document exists, **DO NOT proceed** — go back to Step 2. A plan cannot be created without a discovery document. This is a hard requirement with no exceptions.
|
|
79
|
+
|
|
80
|
+
1. **Verify** discovery document exists in `flow/discovery/` (created in Step 2)
|
|
81
|
+
2. **Read** the file `.claude/commands/create-plan.md` to load planning context
|
|
82
|
+
3. Execute the create-plan skill with the discovery document as input
|
|
83
|
+
4. **PAUSE**: Present the plan for user approval (mandatory checkpoint)
|
|
84
|
+
5. Wait for user to approve the plan before proceeding
|
|
85
|
+
6. Produce the plan document in `flow/plans/`
|
|
86
|
+
7. Write transition summary
|
|
87
|
+
8. **Auto-proceed** to Step 4 (override the "no auto-chaining" rule)
|
|
88
|
+
|
|
89
|
+
### Step 4: Execute Plan
|
|
90
|
+
|
|
91
|
+
1. **Read** the file `.claude/commands/execute-plan.md` to load execution context
|
|
92
|
+
2. Execute the plan following complexity-based grouping strategy
|
|
93
|
+
3. Update the plan file with progress
|
|
94
|
+
4. Run build + test verification at the end
|
|
95
|
+
5. Write transition summary
|
|
96
|
+
6. **Auto-proceed** to Step 5
|
|
97
|
+
|
|
98
|
+
### Step 5: Review Code
|
|
99
|
+
|
|
100
|
+
1. **Read** the file `.claude/commands/review-code.md` to load review context
|
|
101
|
+
2. Review all uncommitted changes
|
|
102
|
+
3. Present the review summary
|
|
103
|
+
|
|
104
|
+
### Step 6: Archive and Complete
|
|
105
|
+
|
|
106
|
+
1. Move the discovery document to `flow/archive/`
|
|
107
|
+
2. Move the plan document to `flow/archive/`
|
|
108
|
+
3. Present completion summary
|
|
109
|
+
4. **Prompt for context cleanup** (see below)
|
|
110
|
+
|
|
111
|
+
---
|
|
112
|
+
|
|
113
|
+
## Transition Summary Format
|
|
114
|
+
|
|
115
|
+
At each step boundary, write this brief summary to bridge context:
|
|
116
|
+
|
|
117
|
+
```
|
|
118
|
+
## Flow Progress
|
|
119
|
+
|
|
120
|
+
**Feature**: [feature name from user input]
|
|
121
|
+
**Current Step**: [N] of 6
|
|
122
|
+
**Completed**:
|
|
123
|
+
- Step 1: Contracts check → [found/none]
|
|
124
|
+
- Step 2: Discovery → `flow/discovery/discovery_<feature>_v1.md`
|
|
125
|
+
- Step 3: Plan → `flow/plans/plan_<feature>_v1.md`
|
|
126
|
+
**Next**: [description of next step]
|
|
127
|
+
**Key context**: [1-2 sentences of what matters for the next step]
|
|
128
|
+
```
|
|
129
|
+
|
|
130
|
+
---
|
|
131
|
+
|
|
132
|
+
## Context Cleanup Prompt
|
|
133
|
+
|
|
134
|
+
After Step 6 completes, always prompt:
|
|
135
|
+
|
|
136
|
+
```
|
|
137
|
+
Flow complete! All artifacts archived.
|
|
138
|
+
|
|
139
|
+
**Summary**:
|
|
140
|
+
- Discovery: archived
|
|
141
|
+
- Plan: archived
|
|
142
|
+
- Code: reviewed
|
|
143
|
+
- Build: passing
|
|
144
|
+
- Tests: passing
|
|
145
|
+
|
|
146
|
+
For best results on your next feature, consider starting a fresh context with `/clear`.
|
|
147
|
+
Ready for the next task?
|
|
148
|
+
```
|
|
149
|
+
|
|
150
|
+
---
|
|
151
|
+
|
|
152
|
+
## Mandatory Checkpoints
|
|
153
|
+
|
|
154
|
+
Even in autopilot, these checkpoints **always** pause for user input:
|
|
155
|
+
|
|
156
|
+
| Checkpoint | When | Why |
|
|
157
|
+
| --- | --- | --- |
|
|
158
|
+
| Discovery Q&A | Step 2 | User must answer requirements questions |
|
|
159
|
+
| Discovery Gate | Step 2→3 | Discovery document must exist before plan creation. No exceptions. |
|
|
160
|
+
| Plan Approval | Step 3 | User must review and approve the plan before execution |
|
|
161
|
+
|
|
162
|
+
All other transitions happen automatically.
|
|
163
|
+
|
|
164
|
+
**Hard Rule**: Step 3 (Create Plan) CANNOT start unless Step 2 (Discovery) produced a document in `flow/discovery/`. If the discovery document does not exist, the flow must loop back to Step 2. A plan without discovery is incomplete and will lead to poor implementations.
|
|
165
|
+
|
|
166
|
+
---
|
|
167
|
+
|
|
168
|
+
## Overriding No Auto-Chaining Rules
|
|
169
|
+
|
|
170
|
+
When autopilot is ON, the following rules in individual commands are **suspended**:
|
|
171
|
+
|
|
172
|
+
- `discovery-plan.md`: "Ask Before Proceeding" and "No Auto-Execution" rules
|
|
173
|
+
- `create-plan.md`: "No Auto-Chaining" and "Complete and Stop" rules
|
|
174
|
+
- `execute-plan.md`: "Complete and Stop" rule
|
|
175
|
+
- `review-code.md`: completion stop behavior
|
|
176
|
+
|
|
177
|
+
These rules remain fully active when autopilot is OFF.
|
|
178
|
+
|
|
179
|
+
---
|
|
180
|
+
|
|
181
|
+
## Error Handling
|
|
182
|
+
|
|
183
|
+
If any step fails:
|
|
184
|
+
|
|
185
|
+
1. Present the error to the user
|
|
186
|
+
2. Ask how to proceed: retry, skip, or abort the flow
|
|
187
|
+
3. If aborted: note which steps completed and which artifacts exist
|
|
188
|
+
4. The flow can be resumed manually using individual slash commands
|
|
@@ -34,9 +34,9 @@ All implementation plans must be created inside `flow/plans/` using snake_case n
|
|
|
34
34
|
|
|
35
35
|
---
|
|
36
36
|
|
|
37
|
-
### 2.
|
|
37
|
+
### 2. Discovery is Required Before Planning
|
|
38
38
|
|
|
39
|
-
|
|
39
|
+
A discovery document in `flow/discovery/` is **required** before creating any plan. Plans cannot be created without a completed discovery document. If no discovery exists, run `/discovery-plan` first. No exceptions. See `.claude/rules/patterns/discovery-patterns.md` for details.
|
|
40
40
|
|
|
41
41
|
---
|
|
42
42
|
|
package/README.md
CHANGED
|
@@ -99,9 +99,12 @@ Copy `skills/plan-flow/` to your project's `skills/plan-flow/`
|
|
|
99
99
|
| `/review-code` | Review local uncommitted changes |
|
|
100
100
|
| `/review-pr` | Review a Pull Request |
|
|
101
101
|
| `/write-tests` | Generate tests for coverage target |
|
|
102
|
+
| `/flow` | Toggle autopilot mode on/off |
|
|
102
103
|
|
|
103
104
|
## Workflow
|
|
104
105
|
|
|
106
|
+
### Manual (default)
|
|
107
|
+
|
|
105
108
|
```
|
|
106
109
|
1. /setup -> Index project patterns (run once)
|
|
107
110
|
2. /discovery-plan -> Gather requirements for a feature
|
|
@@ -110,6 +113,35 @@ Copy `skills/plan-flow/` to your project's `skills/plan-flow/`
|
|
|
110
113
|
5. /review-code -> Review changes before committing
|
|
111
114
|
```
|
|
112
115
|
|
|
116
|
+
### Autopilot Mode
|
|
117
|
+
|
|
118
|
+
Enable autopilot with `/flow -enable` and the full workflow runs automatically for feature requests:
|
|
119
|
+
|
|
120
|
+
```
|
|
121
|
+
You: "Add dark mode support"
|
|
122
|
+
|
|
123
|
+
-> Contracts check (automatic)
|
|
124
|
+
-> Discovery (pauses for your Q&A)
|
|
125
|
+
-> Create plan (pauses for your approval)
|
|
126
|
+
-> Execute plan (automatic)
|
|
127
|
+
-> Review code (automatic)
|
|
128
|
+
-> Archive (automatic)
|
|
129
|
+
```
|
|
130
|
+
|
|
131
|
+
Autopilot classifies every input and only triggers the full flow for feature requests (complexity 3+). Questions, trivial tasks, and slash commands are handled normally.
|
|
132
|
+
|
|
133
|
+
| Usage | Description |
|
|
134
|
+
|-------|-------------|
|
|
135
|
+
| `/flow -enable` | Enable autopilot mode |
|
|
136
|
+
| `/flow -disable` | Disable autopilot mode |
|
|
137
|
+
| `/flow -status` | Check current state |
|
|
138
|
+
|
|
139
|
+
**Mandatory checkpoints** — even in autopilot, the flow always pauses for:
|
|
140
|
+
- **Discovery Q&A**: You answer requirements questions
|
|
141
|
+
- **Plan approval**: You review and approve the plan before execution
|
|
142
|
+
|
|
143
|
+
State is persisted in `flow/.autopilot` (survives session restarts).
|
|
144
|
+
|
|
113
145
|
## Complexity Scoring
|
|
114
146
|
|
|
115
147
|
Every plan phase has a complexity score (0-10):
|
|
@@ -134,7 +166,8 @@ flow/
|
|
|
134
166
|
├── plans/ # Active implementation plans
|
|
135
167
|
├── references/ # Reference materials
|
|
136
168
|
├── reviewed-code/ # Code review documents
|
|
137
|
-
|
|
169
|
+
├── reviewed-pr/ # PR review documents
|
|
170
|
+
└── ledger.md # Persistent project learning journal
|
|
138
171
|
```
|
|
139
172
|
|
|
140
173
|
## Requirements
|
package/package.json
CHANGED
package/rules/core/_index.mdc
CHANGED
|
@@ -9,8 +9,8 @@ alwaysApply: false
|
|
|
9
9
|
|
|
10
10
|
Core rules define the foundational coding standards that apply across the entire project. These include best practices to follow (allowed patterns), anti-patterns to avoid (forbidden patterns), and complexity scoring for implementation planning.
|
|
11
11
|
|
|
12
|
-
**Total Files**:
|
|
13
|
-
**Reference Codes**: COR-AP-1 through COR-
|
|
12
|
+
**Total Files**: 5 files, ~1011 lines
|
|
13
|
+
**Reference Codes**: COR-AP-1 through COR-AM-3
|
|
14
14
|
|
|
15
15
|
---
|
|
16
16
|
|
|
@@ -40,6 +40,14 @@ Core rules define the foundational coding standards that apply across the entire
|
|
|
40
40
|
| COR-CS-4 | Common complexity patterns table | complexity-scoring.mdc | 168-185 |
|
|
41
41
|
| COR-CS-5 | Real-world scoring examples | complexity-scoring.mdc | 187-236 |
|
|
42
42
|
|
|
43
|
+
### Autopilot Mode (`autopilot-mode.mdc`)
|
|
44
|
+
|
|
45
|
+
| Code | Description | Source | Lines |
|
|
46
|
+
|------|-------------|--------|-------|
|
|
47
|
+
| COR-AM-1 | Input classification (slash commands, questions, trivial, feature requests) | autopilot-mode.mdc | 18-50 |
|
|
48
|
+
| COR-AM-2 | Flow execution steps (contracts, discovery, plan, execute, review, archive) | autopilot-mode.mdc | 52-110 |
|
|
49
|
+
| COR-AM-3 | Mandatory checkpoints and overriding rules | autopilot-mode.mdc | 140-178 |
|
|
50
|
+
|
|
43
51
|
### Project Ledger (`project-ledger.mdc`)
|
|
44
52
|
|
|
45
53
|
| Code | Description | Source | Lines |
|
|
@@ -64,6 +72,9 @@ Core rules define the foundational coding standards that apply across the entire
|
|
|
64
72
|
| COR-CS-5 | Need detailed scoring examples |
|
|
65
73
|
| COR-LDG-1 | Need ledger behavior and entry guidelines |
|
|
66
74
|
| COR-LDG-2 | Need ledger maintenance and integration rules |
|
|
75
|
+
| COR-AM-1 | Need to understand input classification for autopilot |
|
|
76
|
+
| COR-AM-2 | Need autopilot flow execution steps |
|
|
77
|
+
| COR-AM-3 | Need mandatory checkpoint and override rules |
|
|
67
78
|
|
|
68
79
|
---
|
|
69
80
|
|
|
@@ -97,3 +108,4 @@ Core rules define the foundational coding standards that apply across the entire
|
|
|
97
108
|
- `allowed-patterns.mdc` and `forbidden-patterns.mdc` have `alwaysApply: true` - they are loaded automatically by Cursor
|
|
98
109
|
- `project-ledger.mdc` has `alwaysApply: true` - maintains persistent project memory
|
|
99
110
|
- `complexity-scoring.mdc` is loaded on-demand when needed for planning
|
|
111
|
+
- `autopilot-mode.mdc` has `alwaysApply: true` - detects and orchestrates autopilot flow mode
|
|
@@ -0,0 +1,193 @@
|
|
|
1
|
+
---
|
|
2
|
+
description: "Autopilot flow mode - when flow/.autopilot exists, automatically orchestrates discovery → plan → execute → review for feature requests"
|
|
3
|
+
alwaysApply: true
|
|
4
|
+
---
|
|
5
|
+
|
|
6
|
+
# Autopilot Flow Mode
|
|
7
|
+
|
|
8
|
+
## Purpose
|
|
9
|
+
|
|
10
|
+
When `flow/.autopilot` exists, this rule is active. It automatically orchestrates the full plan-flow workflow for feature requests, loading only the relevant command context at each step.
|
|
11
|
+
|
|
12
|
+
---
|
|
13
|
+
|
|
14
|
+
## On Session Start
|
|
15
|
+
|
|
16
|
+
1. Check if `flow/.autopilot` exists
|
|
17
|
+
2. If it exists: autopilot mode is **ON** - follow the rules below for every user input
|
|
18
|
+
3. If it does not exist: autopilot mode is **OFF** - ignore this rule entirely, normal behavior applies
|
|
19
|
+
|
|
20
|
+
---
|
|
21
|
+
|
|
22
|
+
## Input Classification
|
|
23
|
+
|
|
24
|
+
When autopilot is ON, classify every new user input before acting:
|
|
25
|
+
|
|
26
|
+
### Category 1: Slash Command
|
|
27
|
+
**Signals**: Input starts with `/` (e.g., `/review-pr`, `/write-tests`, `/flow -disable`)
|
|
28
|
+
**Action**: Run that command normally. Do NOT trigger the flow.
|
|
29
|
+
|
|
30
|
+
### Category 2: Question or Exploration
|
|
31
|
+
**Signals**: User is asking about code, requesting explanations, exploring the codebase, or asking for help.
|
|
32
|
+
**Examples**: "What does this function do?", "How does auth work here?", "Show me the API routes"
|
|
33
|
+
**Action**: Answer normally. Do NOT trigger the flow.
|
|
34
|
+
|
|
35
|
+
### Category 3: Trivial Task
|
|
36
|
+
**Signals**: Single-file change, obvious fix, user specifies exact change, no architectural decisions needed. Estimated complexity 0-2.
|
|
37
|
+
**Examples**: "Fix the typo in README", "Rename this variable to camelCase", "Add a missing import"
|
|
38
|
+
**Action**: Execute directly. Do NOT trigger the flow.
|
|
39
|
+
|
|
40
|
+
### Category 4: Feature Request
|
|
41
|
+
**Signals**: Adding new functionality, behavior changes, multi-file scope, mentions creating/building/implementing, references APIs/integrations/components. Estimated complexity 3+.
|
|
42
|
+
**Examples**: "Add dark mode support", "Implement user authentication", "Create a new API endpoint for payments"
|
|
43
|
+
**Action**: **Trigger the full flow** (see below).
|
|
44
|
+
|
|
45
|
+
### When Uncertain
|
|
46
|
+
If the input doesn't clearly fit a category, use the `AskUserQuestion` tool:
|
|
47
|
+
|
|
48
|
+
```typescript
|
|
49
|
+
AskUserQuestion({
|
|
50
|
+
questions: [{
|
|
51
|
+
question: "I'm in autopilot mode. How should I handle this request?",
|
|
52
|
+
header: "Flow mode",
|
|
53
|
+
options: [
|
|
54
|
+
{ label: "Run full flow", description: "Run the full workflow: discovery → plan → execute → review" },
|
|
55
|
+
{ label: "Handle directly", description: "Just handle this task directly without the full flow" }
|
|
56
|
+
],
|
|
57
|
+
multiSelect: false
|
|
58
|
+
}]
|
|
59
|
+
})
|
|
60
|
+
```
|
|
61
|
+
|
|
62
|
+
---
|
|
63
|
+
|
|
64
|
+
## Flow Execution Steps
|
|
65
|
+
|
|
66
|
+
When a **feature request** is detected, execute these steps in order. At each step, **Read the relevant command file** to load its context, then follow its instructions.
|
|
67
|
+
|
|
68
|
+
### Step 1: Check Contracts
|
|
69
|
+
|
|
70
|
+
Check `flow/contracts/` for any relevant integration contracts. If found, note them for the discovery step.
|
|
71
|
+
|
|
72
|
+
### Step 2: Discovery
|
|
73
|
+
|
|
74
|
+
1. **Read** the discovery command to load discovery context
|
|
75
|
+
2. Execute the discovery skill following that command's instructions
|
|
76
|
+
3. **PAUSE**: Ask user discovery questions as needed (mandatory checkpoint)
|
|
77
|
+
4. Produce the discovery document in `flow/discovery/`
|
|
78
|
+
5. Write transition summary (see format below)
|
|
79
|
+
6. **Auto-proceed** to Step 3 (override the "ask before proceeding" rule)
|
|
80
|
+
|
|
81
|
+
### Step 3: Create Plan
|
|
82
|
+
|
|
83
|
+
**GATE CHECK**: Before proceeding, verify that Step 2 produced a discovery document in `flow/discovery/`. If no discovery document exists, **DO NOT proceed** — go back to Step 2. A plan cannot be created without a discovery document. This is a hard requirement with no exceptions.
|
|
84
|
+
|
|
85
|
+
1. **Verify** discovery document exists in `flow/discovery/` (created in Step 2)
|
|
86
|
+
2. **Read** the create-plan command to load planning context
|
|
87
|
+
3. Execute the create-plan skill with the discovery document as input
|
|
88
|
+
4. **PAUSE**: Present the plan for user approval (mandatory checkpoint)
|
|
89
|
+
5. Wait for user to approve the plan before proceeding
|
|
90
|
+
6. Produce the plan document in `flow/plans/`
|
|
91
|
+
7. Write transition summary
|
|
92
|
+
8. **Auto-proceed** to Step 4 (override the "no auto-chaining" rule)
|
|
93
|
+
|
|
94
|
+
### Step 4: Execute Plan
|
|
95
|
+
|
|
96
|
+
1. **Read** the execute-plan command to load execution context
|
|
97
|
+
2. Execute the plan following complexity-based grouping strategy
|
|
98
|
+
3. Update the plan file with progress
|
|
99
|
+
4. Run build + test verification at the end
|
|
100
|
+
5. Write transition summary
|
|
101
|
+
6. **Auto-proceed** to Step 5
|
|
102
|
+
|
|
103
|
+
### Step 5: Review Code
|
|
104
|
+
|
|
105
|
+
1. **Read** the review-code command to load review context
|
|
106
|
+
2. Review all uncommitted changes
|
|
107
|
+
3. Present the review summary
|
|
108
|
+
|
|
109
|
+
### Step 6: Archive and Complete
|
|
110
|
+
|
|
111
|
+
1. Move the discovery document to `flow/archive/`
|
|
112
|
+
2. Move the plan document to `flow/archive/`
|
|
113
|
+
3. Present completion summary
|
|
114
|
+
4. **Prompt for context cleanup** (see below)
|
|
115
|
+
|
|
116
|
+
---
|
|
117
|
+
|
|
118
|
+
## Transition Summary Format
|
|
119
|
+
|
|
120
|
+
At each step boundary, write this brief summary to bridge context:
|
|
121
|
+
|
|
122
|
+
```
|
|
123
|
+
## Flow Progress
|
|
124
|
+
|
|
125
|
+
**Feature**: [feature name from user input]
|
|
126
|
+
**Current Step**: [N] of 6
|
|
127
|
+
**Completed**:
|
|
128
|
+
- Step 1: Contracts check → [found/none]
|
|
129
|
+
- Step 2: Discovery → `flow/discovery/discovery_<feature>_v1.md`
|
|
130
|
+
- Step 3: Plan → `flow/plans/plan_<feature>_v1.md`
|
|
131
|
+
**Next**: [description of next step]
|
|
132
|
+
**Key context**: [1-2 sentences of what matters for the next step]
|
|
133
|
+
```
|
|
134
|
+
|
|
135
|
+
---
|
|
136
|
+
|
|
137
|
+
## Context Cleanup Prompt
|
|
138
|
+
|
|
139
|
+
After Step 6 completes, always prompt:
|
|
140
|
+
|
|
141
|
+
```
|
|
142
|
+
Flow complete! All artifacts archived.
|
|
143
|
+
|
|
144
|
+
**Summary**:
|
|
145
|
+
- Discovery: archived
|
|
146
|
+
- Plan: archived
|
|
147
|
+
- Code: reviewed
|
|
148
|
+
- Build: passing
|
|
149
|
+
- Tests: passing
|
|
150
|
+
|
|
151
|
+
For best results on your next feature, consider starting a fresh context with `/clear`.
|
|
152
|
+
Ready for the next task?
|
|
153
|
+
```
|
|
154
|
+
|
|
155
|
+
---
|
|
156
|
+
|
|
157
|
+
## Mandatory Checkpoints
|
|
158
|
+
|
|
159
|
+
Even in autopilot, these checkpoints **always** pause for user input:
|
|
160
|
+
|
|
161
|
+
| Checkpoint | When | Why |
|
|
162
|
+
| --- | --- | --- |
|
|
163
|
+
| Discovery Q&A | Step 2 | User must answer requirements questions |
|
|
164
|
+
| Discovery Gate | Step 2→3 | Discovery document must exist before plan creation. No exceptions. |
|
|
165
|
+
| Plan Approval | Step 3 | User must review and approve the plan before execution |
|
|
166
|
+
|
|
167
|
+
All other transitions happen automatically.
|
|
168
|
+
|
|
169
|
+
**Hard Rule**: Step 3 (Create Plan) CANNOT start unless Step 2 (Discovery) produced a document in `flow/discovery/`. If the discovery document does not exist, the flow must loop back to Step 2. A plan without discovery is incomplete and will lead to poor implementations.
|
|
170
|
+
|
|
171
|
+
---
|
|
172
|
+
|
|
173
|
+
## Overriding No Auto-Chaining Rules
|
|
174
|
+
|
|
175
|
+
When autopilot is ON, the following rules in individual commands are **suspended**:
|
|
176
|
+
|
|
177
|
+
- Discovery command: "Ask Before Proceeding" and "No Auto-Execution" rules
|
|
178
|
+
- Create-plan command: "No Auto-Chaining" and "Complete and Stop" rules
|
|
179
|
+
- Execute-plan command: "Complete and Stop" rule
|
|
180
|
+
- Review-code command: completion stop behavior
|
|
181
|
+
|
|
182
|
+
These rules remain fully active when autopilot is OFF.
|
|
183
|
+
|
|
184
|
+
---
|
|
185
|
+
|
|
186
|
+
## Error Handling
|
|
187
|
+
|
|
188
|
+
If any step fails:
|
|
189
|
+
|
|
190
|
+
1. Present the error to the user
|
|
191
|
+
2. Ask how to proceed: retry, skip, or abort the flow
|
|
192
|
+
3. If aborted: note which steps completed and which artifacts exist
|
|
193
|
+
4. The flow can be resumed manually using individual slash commands
|
|
@@ -46,9 +46,9 @@ All implementation plans must be created inside `flow/plans/` using snake_case n
|
|
|
46
46
|
|
|
47
47
|
---
|
|
48
48
|
|
|
49
|
-
### 2.
|
|
49
|
+
### 2. Discovery is Required Before Planning
|
|
50
50
|
|
|
51
|
-
|
|
51
|
+
A discovery document in `flow/discovery/` is **required** before creating any plan. Plans cannot be created without a completed discovery document. If no discovery exists, run `/discovery-plan` first. No exceptions. See `.cursor/rules/patterns/discovery-patterns.mdc` for details.
|
|
52
52
|
|
|
53
53
|
---
|
|
54
54
|
|
|
@@ -57,7 +57,7 @@ This skill is **strictly for creating plan documents**. The process:
|
|
|
57
57
|
|
|
58
58
|
| Input | Required | Description |
|
|
59
59
|
| -------------------- | -------- | -------------------------------------------------- |
|
|
60
|
-
| `discovery_document` |
|
|
60
|
+
| `discovery_document` | **Required** | Path to discovery document. Plans cannot be created without one. |
|
|
61
61
|
| `feature_name` | Yes | Name of the feature to plan |
|
|
62
62
|
| `requirements` | Optional | Direct requirements if no discovery document |
|
|
63
63
|
| `version` | Optional | Version number (auto-incremented if not provided) |
|
|
@@ -66,19 +66,16 @@ This skill is **strictly for creating plan documents**. The process:
|
|
|
66
66
|
|
|
67
67
|
## Workflow
|
|
68
68
|
|
|
69
|
-
### Step 1:
|
|
69
|
+
### Step 1: Validate Discovery Document (HARD BLOCK)
|
|
70
70
|
|
|
71
|
-
**
|
|
71
|
+
**A discovery document is REQUIRED. No exceptions.**
|
|
72
72
|
|
|
73
|
-
1.
|
|
74
|
-
2.
|
|
75
|
-
3.
|
|
76
|
-
|
|
77
|
-
|
|
78
|
-
|
|
79
|
-
|
|
80
|
-
1. Use requirements provided directly
|
|
81
|
-
2. Note that discovery was skipped
|
|
73
|
+
1. Check if a discovery document was provided or exists in `flow/discovery/`
|
|
74
|
+
2. **If NO discovery document exists**: **STOP immediately.** Do NOT create a plan. Inform the user that discovery is required and run `/discovery-plan` first.
|
|
75
|
+
3. **If discovery document exists**: Read it and extract:
|
|
76
|
+
- Feature name, description, goals
|
|
77
|
+
- FR, NFR, Constraints
|
|
78
|
+
- Risks identified
|
|
82
79
|
|
|
83
80
|
---
|
|
84
81
|
|
|
@@ -254,7 +251,7 @@ Before completing the plan, verify:
|
|
|
254
251
|
- [ ] All phases have complexity scores (X/10)
|
|
255
252
|
- [ ] Tests are the LAST phase
|
|
256
253
|
- [ ] Key Changes section is populated
|
|
257
|
-
- [ ] Discovery document is referenced (
|
|
254
|
+
- [ ] Discovery document is referenced (discovery is required, never skipped)
|
|
258
255
|
- [ ] **NO implementation code is included**
|
|
259
256
|
- [ ] **NO source files were created or modified**
|
|
260
257
|
|
|
@@ -22,6 +22,7 @@ A comprehensive skill set for AI-assisted software development with structured w
|
|
|
22
22
|
| `/review-code` | Review local uncommitted changes |
|
|
23
23
|
| `/review-pr` | Review a Pull Request |
|
|
24
24
|
| `/write-tests` | Write tests to achieve coverage target |
|
|
25
|
+
| `/flow` | Toggle autopilot mode (auto-chains the full workflow) |
|
|
25
26
|
|
|
26
27
|
## Always-Active Features
|
|
27
28
|
|
|
@@ -83,7 +84,7 @@ flow/
|
|
|
83
84
|
## Critical Rules
|
|
84
85
|
|
|
85
86
|
1. **Automated Workflow**: Run discovery → plan → execute automatically. Only stop to ask when you need information from the user.
|
|
86
|
-
2. **Discovery First
|
|
87
|
+
2. **Discovery First (Hard Block)**: `/discovery` is **required** before `/create-plan`. Plans cannot be created without a discovery document. No exceptions. If no discovery exists, run discovery first.
|
|
87
88
|
3. **Tests Last**: Tests are always the last phase of any implementation plan.
|
|
88
89
|
4. **Build at End Only**: Run build verification only after ALL phases complete.
|
|
89
90
|
5. **Archive When Done**: Move completed discovery and plans to `flow/archive/`.
|
|
@@ -0,0 +1,81 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: flow
|
|
3
|
+
description: Toggle autopilot flow mode - automatically orchestrates discovery, planning, execution, and review for feature requests
|
|
4
|
+
metadata: {"openclaw":{"requires":{"bins":["git"]}}}
|
|
5
|
+
user-invocable: true
|
|
6
|
+
---
|
|
7
|
+
|
|
8
|
+
# Flow: Autopilot Mode Toggle
|
|
9
|
+
|
|
10
|
+
Enables or disables autopilot flow mode. When enabled, feature requests automatically run the full plan-flow workflow without manual command invocation.
|
|
11
|
+
|
|
12
|
+
## Usage
|
|
13
|
+
|
|
14
|
+
```
|
|
15
|
+
/flow -enable Enable autopilot mode
|
|
16
|
+
/flow -disable Disable autopilot mode
|
|
17
|
+
/flow -status Show current state
|
|
18
|
+
/flow Same as -status
|
|
19
|
+
/flow -help Show help
|
|
20
|
+
```
|
|
21
|
+
|
|
22
|
+
## How It Works
|
|
23
|
+
|
|
24
|
+
**State**: Stored in `flow/.autopilot` (empty marker file). Create to enable, delete to disable.
|
|
25
|
+
|
|
26
|
+
### When Enabled
|
|
27
|
+
|
|
28
|
+
Every user input is classified:
|
|
29
|
+
|
|
30
|
+
| Category | Signals | Action |
|
|
31
|
+
|----------|---------|--------|
|
|
32
|
+
| Slash Command | Starts with `/` | Run normally, no flow |
|
|
33
|
+
| Question/Exploration | Asking about code, explanations | Answer normally, no flow |
|
|
34
|
+
| Trivial Task | Single-file, obvious fix, complexity 0-2 | Execute directly, no flow |
|
|
35
|
+
| Feature Request | New functionality, multi-file, complexity 3+ | **Trigger full flow** |
|
|
36
|
+
|
|
37
|
+
### Full Flow (6 Steps)
|
|
38
|
+
|
|
39
|
+
1. **Check Contracts** → Look for relevant integration contracts
|
|
40
|
+
2. **Discovery** (PAUSE) → Gather requirements, ask user questions
|
|
41
|
+
3. **Create Plan** (PAUSE) → Create implementation plan, get user approval
|
|
42
|
+
4. **Execute Plan** → Implement phases with complexity-based grouping
|
|
43
|
+
5. **Review Code** → Review all uncommitted changes
|
|
44
|
+
6. **Archive** → Move artifacts to `flow/archive/`, prompt for cleanup
|
|
45
|
+
|
|
46
|
+
### Mandatory Checkpoints
|
|
47
|
+
|
|
48
|
+
Even in autopilot, these always pause for user input:
|
|
49
|
+
|
|
50
|
+
| Checkpoint | When | Why |
|
|
51
|
+
|------------|------|-----|
|
|
52
|
+
| Discovery Q&A | Step 2 | User must answer requirements questions |
|
|
53
|
+
| Discovery Gate | Step 2→3 | Discovery document must exist before plan. No exceptions. |
|
|
54
|
+
| Plan Approval | Step 3 | User must approve plan before execution |
|
|
55
|
+
|
|
56
|
+
### When Disabled
|
|
57
|
+
|
|
58
|
+
Commands run individually as normal:
|
|
59
|
+
`/discovery` → `/create-plan` → `/execute-plan` → `/review-code`
|
|
60
|
+
|
|
61
|
+
## Toggle Commands
|
|
62
|
+
|
|
63
|
+
### Enable
|
|
64
|
+
1. Create `flow/.autopilot` (empty file)
|
|
65
|
+
2. Confirm to user that autopilot is enabled
|
|
66
|
+
|
|
67
|
+
### Disable
|
|
68
|
+
1. Delete `flow/.autopilot`
|
|
69
|
+
2. Confirm to user that autopilot is disabled
|
|
70
|
+
|
|
71
|
+
### Status
|
|
72
|
+
1. Check if `flow/.autopilot` exists
|
|
73
|
+
2. Report current state (ENABLED/DISABLED)
|
|
74
|
+
|
|
75
|
+
## Critical Rules
|
|
76
|
+
|
|
77
|
+
| Rule | Description |
|
|
78
|
+
|------|-------------|
|
|
79
|
+
| **File-based state** | State is stored in `flow/.autopilot` - create to enable, delete to disable |
|
|
80
|
+
| **No other side effects** | This command ONLY manages the marker file. It does not run any workflow steps. |
|
|
81
|
+
| **Complete and stop** | After toggling state, STOP and wait for user input |
|
|
@@ -12,6 +12,7 @@
|
|
|
12
12
|
| `/review-code` | Review local uncommitted changes |
|
|
13
13
|
| `/review-pr` | Review a Pull Request |
|
|
14
14
|
| `/write-tests` | Write tests to achieve coverage target |
|
|
15
|
+
| `/flow` | Toggle autopilot mode (auto-chains discovery → plan → execute → review) |
|
|
15
16
|
|
|
16
17
|
## Recommended Workflow
|
|
17
18
|
|
|
@@ -27,7 +28,7 @@ The Project Ledger (`flow/ledger.md`) is always active. It silently captures pro
|
|
|
27
28
|
|
|
28
29
|
## Critical Rules
|
|
29
30
|
|
|
30
|
-
1. **No Auto-Chaining**: Never auto-invoke the next command. Wait for user.
|
|
31
|
+
1. **No Auto-Chaining**: Never auto-invoke the next command. Wait for user. **Exception**: When autopilot mode is ON (`flow/.autopilot` exists), commands auto-chain per the autopilot flow skill.
|
|
31
32
|
2. **Discovery First**: Recommend `/discovery-plan` before `/create-plan` for new features.
|
|
32
33
|
3. **Tests Last**: Tests are always the last phase of any implementation plan.
|
|
33
34
|
4. **Build at End Only**: Run `npm run build` only after ALL phases complete.
|
|
@@ -12,6 +12,7 @@
|
|
|
12
12
|
| `/review-code` | Review local uncommitted changes |
|
|
13
13
|
| `/review-pr` | Review a Pull Request |
|
|
14
14
|
| `/write-tests` | Write tests to achieve coverage target |
|
|
15
|
+
| `/flow` | Toggle autopilot mode (auto-chains discovery → plan → execute → review) |
|
|
15
16
|
|
|
16
17
|
## Recommended Workflow
|
|
17
18
|
|
|
@@ -27,8 +28,8 @@ The Project Ledger (`flow/ledger.md`) is always active. It silently captures pro
|
|
|
27
28
|
|
|
28
29
|
## Critical Rules
|
|
29
30
|
|
|
30
|
-
1. **No Auto-Chaining**: Never auto-invoke the next command. Wait for user.
|
|
31
|
-
2. **Discovery First**:
|
|
31
|
+
1. **No Auto-Chaining**: Never auto-invoke the next command. Wait for user. **Exception**: When autopilot mode is ON (`flow/.autopilot` exists), commands auto-chain per `.claude/rules/core/autopilot-mode.md`.
|
|
32
|
+
2. **Discovery First**: `/discovery-plan` is **required** before `/create-plan`. Plans cannot be created without a discovery document. No exceptions.
|
|
32
33
|
3. **Tests Last**: Tests are always the last phase of any implementation plan.
|
|
33
34
|
4. **Build at End Only**: Run `npm run build` only after ALL phases complete.
|
|
34
35
|
|