home/lifestyle/prompt-evolution-tree

Prompt Evolution Tree

GPTClaudeGemini··1,181 copies·updated 2026-07-14
prompt-evolution-tree.prompt
# Prompt Evolution Tree (PET) Schema

The Prompt Evolution Tree (PET) is a **directed graph** of prompts and their variations, stored in the `nodes` sheet of the ledger.

Each **node** represents:

- A prompt (or meta-prompt)
- Its type and role in the system
- Links to descriptive artifacts (Docs, assets)
- Relationships to parent / ancestor prompts

Each **run** represents:

- A single execution of a node’s prompt
- The model used, context, and evaluation data

---

## 🌲 Node Schema (logical)

Each node has:

- `node_id` — globally unique identifier (e.g. `PET-00001`)
- `parent_node_id` — optional; `""` or `ROOT` if top-level
- `title` — human-readable name of the prompt
- `prompt_type` — category (e.g. `workspace_automation`, `devlog`, `summarizer`)
- `doc_url` — link to Google Doc that describes the prompt
- `status` — e.g. `draft`, `active`, `archived`
- `tags` — comma-separated labels (e.g. `gmail,summary,v1`)
- `created_at` — ISO datetime
- `updated_at` — ISO datetime

These fields are stored as columns in the `nodes` sheet.

---

## 🧪 Run Schema (logical)

Each run has:

- `run_id` — unique identifier for this execution
- `node_id` — foreign key → which node was run
- `model` — model identifier (e.g. `gemini-1.5-pro`)
- `run_context` — surface or app (e.g. `gmail`, `docs`, `forms`, `cli`)
- `input_ref` — reference to input payload (URL, ID, or short label)
- `output_ref` — reference to result (URL, ID, or summary label)
- `score` — numeric evaluation (0–1, 0–100, etc.)
- `notes` — free-form notes
- `run_time` — ISO datetime

These fields are stored as columns in the `runs` sheet.

---

## 🧾 JSON Example

Example JSON node (also represented in `examples/prompt_tree_example.json`):

when to use it

Community prompt sourced from the open-source GitHub repo PEACEBINFLOW/mindseye-google-ledger (MIT). A "Prompt Evolution Tree" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.

tags

lifestylecommunitygeneral

source

PEACEBINFLOW/mindseye-google-ledger · MIT