Skip to content
Get packs

Guide · Agent skill

How to use the Prompting agent skill: install, trigger and examples

Updated · written with AI from the project's own documentation and checked against our catalogue data

The Prompting skill is a meta-prompting library for AI agents. It helps an agent write, optimize and compose the instructions you give other AI systems, such as system prompts, agent briefs and reusable templates. This guide explains what the skill does, how to install it, how to trigger it with plain requests, and what it needs to run, so you can use it without prior experience with agent skills.

Prompt Writing Librarydanielmiessler/LifeOS · MIT

Quick start

  1. 1Download the ZIP. Download the Prompting skill ZIP from this page.
  2. 2Unzip into a skills folder. Unzip into ~/.claude/skills/ for personal use, or into .claude/skills/ in a project.
  3. 3Or upload in the Claude apps. Under Settings → Capabilities → Skills, upload the skill if you use the Claude apps.
  4. 4Install template packages if needed. Run cd ~/.claude && bun install so the handlebars and yaml packages are available to the template tools.
  5. 5Ask for a prompt in plain language. Request a prompt, system prompt or template in natural words, naming the audience and constraints.
  6. 6Check the output. Review the generated prompt against your outcomes and constraints, and test it before relying on it.

What the Prompting skill makes the agent do

The skill gives the agent a standard set of rules and tools for writing prompts. When it is installed, the agent aims to describe the desired outcome, the constraints and the output format, rather than a step-by-step procedure. The skill's own standard says the only procedural lines that should survive are safety gates, verified gotchas, tool contracts and output-format contracts.

The skill can also fill reusable Handlebars templates with data through a bundled render script. Structure stays in the template and content comes from a YAML or JSON data file.

The output is always a prompt to use somewhere else. It is not the finished article, code or report. For that work, the skill points to a domain skill.

Install the Prompting skill

Download the ZIP from this page. Then choose one of two routes.

For a personal install, unzip it into ~/.claude/skills/. For a project-only install, unzip it into the project's .claude/skills/ folder.

In the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

  • Personal: ~/.claude/skills/
  • Project: .claude/skills/ inside your project
  • Claude apps: Settings → Capabilities → Skills

Set up the template tools (only if you need them)

The standards and reference documents work by reading alone. Only the two template tools need extra setup. Those tools use the Bun runtime and the handlebars and yaml npm packages.

The setup notes say both packages are declared in the root package.json and arrive with bun install. If you skipped that step, run cd ~/.claude && bun install. A quick check is to run bun ~/.claude/skills/Prompting/Tools/RenderTemplate.ts --help.

  • Requires the Bun runtime
  • Requires the handlebars and yaml npm packages
  • Not needed for standards-only use
cd ~/.claude && bun install

Trigger the skill with natural requests

You do not need a special command. Ask in plain language about prompts, and the agent should pick up the skill. The skill's trigger phrases include meta-prompting, template generation, prompt optimization, prompt engineering, write a prompt, create system prompt and Handlebars template.

Be specific about the output you want. Name the prompt type, the audience (such as another agent), and the constraints that matter.

  • Write a prompt that briefs a research agent on reviewing a security design.
  • Optimize this system prompt so it states the outcome and constraints instead of steps.
  • Create a system prompt for a support bot that must refuse billing changes.
  • Render this Handlebars template with my agents.yaml data.

Render a template from data

Templates live under Templates/. Pass a template path and a data object, or pass a .yaml or .json file. Relative paths are looked up under Templates/ first, then in the current directory.

The example below renders a briefing template from a data file and prints a preview.

bun run ${LIFEOS_SKILL_DIR}/Tools/RenderTemplate.ts --template Primitives/Briefing.hbs --data attack-surface-brief.yaml --preview

Write an ideal-state data file

Keep the data to outcomes, constraints and available tools, with no steps. The skill's example puts the task deliverables as testable outcomes, lists read-only limits as constraints, and sets the output type to markdown.

Check the result before you rely on it. The skill says a generated prompt should be tested before it is declared ready, since a prompt that looks good may perform poorly.

  • State the outcome as testable deliverables
  • List limits as constraints
  • Name the tools the agent may use
  • Set an output format

Safety notes: scripts and network

The skill runs scripts. The template tools are TypeScript files run with Bun, so read them before running them on sensitive data. The skill's listed requirements say it does not use the network.

Keep your data files free of secrets you do not want in a prompt, because the rendered output is text meant to be pasted or passed to other systems.

  • Runs scripts: yes (Bun)
  • Uses network: no
  • Review Tools/RenderTemplate.ts and Tools/ValidateTemplate.ts before running

What people use it for

  • Drafting a system prompt for a developer's agent with a consistent structure.
  • Maintaining a library of reusable prompt templates driven by YAML or JSON data.
  • Writing an agent brief that defines outcomes, constraints and tools for a research or review task.
  • Auditing existing prompts against the skill's ideal-state standard.

Frequently asked questions

Is the Prompting skill free to use?

The skill is released under the MIT licence, as stated on this page. Check the licence file in the source repository for the full terms.

What do I need to run the Prompting skill?

The standards and references need only reading. The template tools need the Bun runtime plus the handlebars and yaml npm packages. Run cd ~/.claude && bun install if you skipped that step.

Where do I install the Prompting skill?

Unzip the downloaded ZIP into ~/.claude/skills/ for personal use or into .claude/skills/ in a project. In the Claude apps, upload it under Settings → Capabilities → Skills.

Does the Prompting skill write final content like articles or code?

No. The skill produces prompts to be used elsewhere, not finished content. For articles, code or reports, use the appropriate domain skill.

Can the Prompting skill grade or evaluate outputs?

No. Judge prompts, rubrics, scoring and grading are meant to run through a separate eval system called Assay, not through this skill.

Why does the template tool fail, and what should I check?

If the tools fail, first confirm Bun is installed and the handlebars and yaml packages are present. Then run the --help check on RenderTemplate.ts to confirm the tool starts.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use the Prompting skill if you write system prompts or agent briefs and want consistent structure, or if you maintain reusable templates driven by YAML or JSON. Pick another option if you need final articles, code or reports, if you need output grading, or if you want zero setup, since the template tools require the handlebars package first.

Download the skill