A privacy-first Chrome extension that intercepts prompts on ChatGPT and Claude, guides you through a structured enhancement framework, and injects the improved prompt before sending. Zero external requests — everything stays on your machine.
PromptForge auto-classifies your prompt and picks the best framework. You can always switch manually inside the modal.
| Framework | Stands For | Best For |
|---|---|---|
| RTF | Role / Task / Format | Short, everyday prompts |
| TAG | Task / Action / Goal | Quick task-oriented prompts |
| RACE | Role / Action / Context / Expectation | Prompts needing context |
| Framework | Stands For | Best For |
|---|---|---|
| CREATE | Character / Request / Examples / Adjustments / Types / Evaluation | Long, complex prompts (PMI Workbook 2024) |
| COSTAR | Context / Objective / Style / Tone / Audience / Response | Audience-aware content |
| RISEN | Role / Instructions / Steps / End Goal / Narrowing | Step-by-step process prompts |
| Framework | Stands For | Best For |
|---|---|---|
| CoT | Chain of Thought | Reasoning, analysis, problem-solving |
| Few-Shot | Examples + Pattern | Matching a specific format or style |
| TRACE | Task / Role / Action / Context / Example | Structured multi-part prompts |
| Framework | Stands For | Best For |
|---|---|---|
| SMART | Specific / Measurable / Achievable / Relevant / Time-bound | Goal-setting, deadline-driven prompts |
| PACT | Purpose / Action / Context / Type | Outcome-focused prompts |
| COAST | Context / Objective / Actions / Scenario / Task | Scenario-based instructions |
- Auto-classifier — detects prompt complexity and selects the best framework
- ? Help-choose panel — 12-item decision guide to pick the right framework
- Manual override — switch framework inside the modal at any time; toggle row is scrollable for all 12
- Framework reason line — shows why a framework was chosen, with a one-liner tip
- Collapsible sections — expand only what you need
- Prompt quality checklist — 8-point PMI self-check with red → amber → green score bar; warns before sending a low-score prompt
- AI Assist — assembles your prompt into a meta-prompt and copies it to clipboard (no API calls)
- Editable preview — review and tweak the assembled prompt before sending
- Save & Download — save prompt pairs to local storage or download as
.mdfiles - History — search, copy, download, and delete saved prompts from the options page
- First-run banner — choose Use PromptForge / Send as-is / Disable on first interception
- Per-framework toggles — enable or disable any of the 12 frameworks independently (grouped by category in options)
- Privacy-first —
chrome.storage.localonly; zero network requests; no telemetry
chatgpt.comchat.openai.comclaude.ai
- Open
chrome://extensions - Enable Developer mode (top right)
- Click Load unpacked
- Select this folder
Open the toolbar popup and click ⚙ to access full settings:
| Setting | Default |
|---|---|
| PromptForge enabled | On |
| Word count threshold | 30 words |
| Default framework | Auto (select dropdown) |
| Per-framework toggles (all 12) | On |
| Show quality checklist | On |
Frameworks in options are grouped by category: Simple / Scenario / Complex / Goal+Technique.
All data is stored in chrome.storage.local — sandboxed to this extension on this machine only. PromptForge makes zero external network requests. Permissions used: storage, activeTab, scripting.
- PMI Prompt Engineering Workbook — "Talking to AI: Prompt Engineering for Project Managers" (2024)
- PMI GenAI Tools for Project Work v1.2 (2023)
- PMI ChatGPT Prompt Engineering Guide (2023)