PromptKit prompt
Microsoft library of personas, protocols, formats and templates from which an agent assembles a prompt for an engineering task.
- 114 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The library consists of text components, and the `promptkit` utility only reads and shows them.
- Interactive mode launches the installed model CLI and gives it a temporary folder through `--add-dir`; the CLI itself decides permissions for project files.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add prompts/promptkitDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The library consists of text components, and the `promptkit` utility only reads and shows them.
- Interactive mode launches the installed model CLI and gives it a temporary folder through `--add-dir`; the CLI itself decides permissions for project files.
This is third-party code. Review the repository files before installing.
What it does
PromptKit stores prompts as versioned components: 15 personas, 48 reasoning protocols, 21 output formats, 5 taxonomies and 64 task templates. The agent reads the manifest, asks what needs doing, picks a persona, protocols and a format, and assembles a finished prompt for a fresh session. The set covers code, firmware, communication protocols, electronics and mechanics. Three interactive workflows form a cycle: `spec-extraction-workflow` extracts requirements, design and validation from existing code; `engineering-workflow` carries a change through specifications with adversarial audits; `maintenance-workflow` finds drift between code and specifications. The `promptkit` utility lists, searches and shows components.
Who it is for. Engineers and tech leads whose teams write prompts for reviews, bug investigations and specifications as one-off notes and want repeatability.
Good fit when
- You need a prompt for a C++ or Rust code review with a memory-safety protocol.
- Requirements and design have to be recovered for an old project without documents.
- A requirements change must pass through specifications and code with a consistency check.
- A team wants to keep prompts in the repository and improve them like code.
Not a fit when
- You need one short hint for an everyday question.
- You have none of Copilot CLI, Claude Code or Codex at hand and would rather not paste files by hand.
- You want a ready service with an interface: the library works inside a model session.
Example request
Assemble a prompt to review `src/network` for data races: systems-engineer persona, a threading protocol and a findings report format.Limitations
Interactive mode needs Node.js 18 or newer and one of the CLIs: GitHub Copilot, Claude Code or Codex. Claude Code does not support skills from the CLI, so launch is manual through `claude "Read and execute bootstrap.md"` in the repository folder. Without a CLI you paste components into a chat yourself following the agent's hints. PromptKit assembles a prompt and does not guarantee the quality of the model's answer. Most materials are written for engineering tasks, with almost no everyday scenarios.
How to disable. Running through `npx` installs nothing permanently, so just stop calling the command. Delete the cloned `promptkit` folder by hand, along with prompts you saved.
FAQ
How do I launch PromptKit?
Run `npx @alan-jowett/promptkit`. The utility looks for an installed model CLI in the order Copilot, Claude Code, Codex and starts a selection session; the `--cli` flag picks another. `npx @alan-jowett/promptkit list --all` prints the components.
How do I use PromptKit with Claude Code?
Clone the repository, enter the folder and run `claude "Read and execute bootstrap.md"`. Codex follows the same scheme with the `codex` command.
How do the three workflows differ?
`spec-extraction-workflow` builds specifications from code; `engineering-workflow` propagates a requirements change with audits; `maintenance-workflow` looks for drift. All three are interactive and run inside a session.
How do I find the right component?
`promptkit search <word>` searches names and descriptions, and `promptkit show <name>` prints a component and the templates that reference it.
Does the kit fit work outside programming?
Yes, it has personas and protocols for electronics, RF, mechanics and standards, but most templates remain engineering ones.
Discussion
No questions or reviews yet.
Related
PromptsA skill that scores a prompt on five criteria, picks one of 31 frameworks and rewrites the request after clarifying questions
A skill that writes precise prompts tailored to a specific tool: Cursor, Claude Code, Midjourney, Stable Diffusion and more
RePrompter interviews you about a rough request and turns it into a scored, structured prompt or a build prompt for a project.
An MCP server and web tool for improving prompts: three tools for optimizing system and user prompts with iterative refinement
Prompt in-context learning
prompt-in-context-learning
An open guide to in-context learning and prompt engineering from EgoAlpha Lab: a collection of papers, code and tools, with part of the material from 2023
A CLI compiler that takes rules, skills, agents and hooks written once and lays them out as native files for Claude Code, Cursor, Copilot and 47 more tools
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| PromptKit (this card) | no | 114 | |
| PromptScript | no | 344 | compiles rules, skills and hooks into files for dozens of agents; PromptKit assembles a finished prompt from layers for a specific engineering task |
| Awesome Prompt Engineering | no | 6,447 | collects links to prompting articles, courses and tools; PromptKit supplies executable components and workflows with audits |
README in short
The README explains the idea, the five component layers and how they are composed, shows a sample session and the utility reference. It also describes the four chains in the manifest and the three end-to-end workflows from extracting specifications to maintenance.