Flint instantly inserts structured, headings-only prompt templates into your editor. No fluff, no token bloat, just lightning-fast prompt structures.
🎥 Video Tutorial
Watch how to use Flint Prompts:
⚡ The Problem & What Flint Solves
- Unstructured Prompts: LLMs output low-quality results without a clear structure (Goal, Context, Edge Cases, Output).
- Token Waste: Copy-pasting verbose prompt templates wastes hundreds of input tokens on explanation boilerplate.
- Flint's Solution: Flint inserts headings-only prompt skeletons that use minimum tokens and force the AI to focus entirely on your code context.
🚀 How to Use (Simple & Fast)
Flint works in any file type (.md, .txt, .ts, .py, etc.) with two quick methods:
Method 1: Autocomplete Slash Commands (Recommended)
- Open any file.
- Type
/followed by the prompt prefix (e.g./feat,/dbg,/arch). - Press Enter or Tab to instantly insert the template structure.
💡 Keyboard Shortcuts:
- Trigger Suggestions Manually: Press
Ctrl+Spaceif the dropdown does not pop up automatically. - Navigate Templates: Use the Up and Down arrow keys to view prompt documentation in the autocomplete hover card.
🔧 Installation Methods
For Visual Studio Code
- Search in UI: Open the Extensions sidebar (
Cmd+Shift+X), search forFlint Prompts, and click Install. - Via Terminal Command:
code --install-extension kamalpulaparthi.promptspec
For Antigravity IDE
- Search in UI: Open the Extensions sidebar inside Antigravity IDE, search for
Flint Prompts, and click Install. - Via Terminal Command:
"/Applications/Antigravity IDE.app/Contents/Resources/app/bin/antigravity-ide" --install-extension kamalpulaparthi.promptspec
📝 Available Prompts & Skeletons
| Slash Trigger | Template Name | Heading Skeleton Inserted |
|---|---|---|
/ai |
General AI | Goal, Context, Constraints, Input, Expected Output |
/dbg |
Debugging | Issue, Expected, Actual, Last Working, Hypotheses, Affected Files, Logs, Tested |
/feat |
Feature Dev | Goal, Stack, Current Architecture, Constraints, Inputs, Edge Cases, Expected Output |
/review |
Code Review | Review checklist: bugs, security, scalability, maintainability, edge cases, complexity |
/arch |
Architecture | System Goal, Limitations, Proposed Arch, Scalability, Security, Trade-offs, Benefits |
/api |
API Design | Goal, Endpoint, Method, Request Body, Expected Response, Validation, Auth, Error Cases |
/refactor |
Refactoring | Goal, Current Problems, Technical Debt, Constraints, Files Involved, Expected Improvements |
/sec |
Security Review | Security audit: auth, authorization, injection, exposure, config, validation, rate limits |
/perf |
Optimization | Goal, Current Bottleneck, Metrics, Affected Modules, Hypotheses, Targets, Improvements |
/agent |
AI Agent | Agent Role, Primary Objective, Tools, Constraints, Decision Rules, Failures, Success Criteria |
/learn |
Learning | Topic, Current Understanding, Knowledge Gaps, Real-World Apps, Exercises, Outcomes |
/decision |
Decision-Making | Decision, Options, Known Facts, Assumptions, Risks, Second-Order Effects, Recommendation |
/rca |
Root Cause | Problem, Start Time, What Changed, Observed Symptoms, Possible Causes, Fix Plan |
/sys |
System Design | Requirements, Traffic, Database, API, Caching, Load Balancing, Failures, Scalability Plan |
/research |
Research | Topic, Objective, Known Facts, Unknowns, Hypotheses, Risk of Assumptions, Expected Insights |
/improve |
Prompt Tuning | Current Prompt, Current Problems, Desired Improvements, Missing Context, Target Quality |
/biz |
Business Logic | Goal, User Flow, Business Rules, Edge Cases, Failure Scenarios, Data Requirements, Output |
/ui |
UI / UX | Goal, Target Users, User Problems, Responsive Rules, A11y Rules, Design Constraints, Structure |
/deploy |
Deployment | Environment, Deployment Target, Error, Last Success, Infrastructure, Logs, Tested |
/context |
Context Eng. | Project Type, Persistent Context Files, Temporary Context, Constraints, Reusable Patterns |
🛡️ Design Principles
- Fully Offline: 100% offline, zero network requests, no AI API keys required.
- Extremely Lightweight: Built purely on the native VS Code Completion and Snippet APIs (no memory-heavy webviews).
📜 Changelog
See the full changelog in the CHANGELOG.md file.
License
MIT