@umgbhalla/pi-gigaplan 0.1.0
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.
- package/README.md +73 -0
- package/extensions/index.ts +847 -0
- package/package.json +53 -0
- package/skills/gigaplan/SKILL.md +60 -0
- package/src/core.ts +720 -0
- package/src/evaluation.ts +383 -0
- package/src/prompts.ts +367 -0
- package/src/schemas.ts +133 -0
- package/src/workers.ts +258 -0
package/README.md
ADDED
|
@@ -0,0 +1,73 @@
|
|
|
1
|
+
# pi-gigaplan
|
|
2
|
+
|
|
3
|
+
Structured AI planning with cross-model critique — a native [pi](https://github.com/badlogic/pi) extension.
|
|
4
|
+
|
|
5
|
+
## What it does
|
|
6
|
+
|
|
7
|
+
Gigaplan coordinates multiple AI agents through a rigorous planning loop:
|
|
8
|
+
|
|
9
|
+
1. **Clarify** — Agent clarifies ambiguous intent
|
|
10
|
+
2. **Plan** — Agent produces a concrete implementation plan
|
|
11
|
+
3. **Critique** — *Different* agent independently raises flags
|
|
12
|
+
4. **Evaluate** — Decision engine scores flags → CONTINUE/SKIP/ESCALATE/ABORT
|
|
13
|
+
5. **Integrate** — Planner addresses flags, revises plan (loops back to critique)
|
|
14
|
+
6. **Gate** — Preflight checks before execution
|
|
15
|
+
7. **Execute** — Agent implements the approved plan
|
|
16
|
+
8. **Review** — Agent validates against success criteria
|
|
17
|
+
|
|
18
|
+
Each step runs as an autonomous **subagent** in a visible cmux terminal pane. You can watch agents work in real-time.
|
|
19
|
+
|
|
20
|
+
## Install
|
|
21
|
+
|
|
22
|
+
```bash
|
|
23
|
+
pi install git:github.com/umgbhalla/pi-gigaplan
|
|
24
|
+
```
|
|
25
|
+
|
|
26
|
+
## Usage
|
|
27
|
+
|
|
28
|
+
```
|
|
29
|
+
/gigaplan build a rate limiter for the API endpoints
|
|
30
|
+
```
|
|
31
|
+
|
|
32
|
+
This will:
|
|
33
|
+
1. Ask for robustness level (light/standard/thorough) and auto-approve preference
|
|
34
|
+
2. Initialize a `.gigaplan/` directory with plan artifacts
|
|
35
|
+
3. Enter gigaplan mode — orchestrating subagents through the full loop
|
|
36
|
+
4. Pause at gate for your approval (unless auto-approve)
|
|
37
|
+
5. Execute and review
|
|
38
|
+
|
|
39
|
+
## Tools
|
|
40
|
+
|
|
41
|
+
| Tool | Description |
|
|
42
|
+
|------|-------------|
|
|
43
|
+
| `gigaplan_step` | Get subagent config for a step |
|
|
44
|
+
| `gigaplan_advance` | Process output and advance state machine |
|
|
45
|
+
| `gigaplan_status` | Show plan status |
|
|
46
|
+
| `gigaplan_override` | Manual intervention (add-note, abort, force-proceed, skip) |
|
|
47
|
+
|
|
48
|
+
## Artifacts
|
|
49
|
+
|
|
50
|
+
All state lives in `.gigaplan/plans/<plan-name>/`:
|
|
51
|
+
|
|
52
|
+
```
|
|
53
|
+
.gigaplan/plans/rate-limiter/
|
|
54
|
+
├── state.json # Mutable plan state
|
|
55
|
+
├── faults.json # Flag registry
|
|
56
|
+
├── plan_v1.md # Versioned plan (markdown)
|
|
57
|
+
├── plan_v1.meta.json # Plan metadata (criteria, assumptions)
|
|
58
|
+
├── critique_v1.json # Critique flags
|
|
59
|
+
├── evaluation_v1.json # Decision engine output
|
|
60
|
+
├── gate.json # Gate preflight results
|
|
61
|
+
├── execution.json # Execution output
|
|
62
|
+
└── review.json # Review results
|
|
63
|
+
```
|
|
64
|
+
|
|
65
|
+
## Configuration
|
|
66
|
+
|
|
67
|
+
Robustness levels control how strict the critique is:
|
|
68
|
+
|
|
69
|
+
| Level | Behavior |
|
|
70
|
+
|-------|----------|
|
|
71
|
+
| **light** | Pragmatic. Only flags real failures. |
|
|
72
|
+
| **standard** | Balanced judgment. Significant risks flagged. |
|
|
73
|
+
| **thorough** | Exhaustive. Edge cases, performance, production concerns. |
|