Prompt Organizer
Open app

Pro feature · Local-first · Portable Markdown

Package the workflow. Not just the prompt.

A prompt answers one question. A SKILL answers a class of questions, the same way, every time. Bundle your instructions, examples, references, and output spec into one portable artifact. Agent Skills (SKILL.md) is an open standard read by Claude Code, Cursor, Copilot CLI, Gemini CLI, and Codex CLI. The format is plain Markdown.

SKILLs export as a directory you own, linted before they ship. No vendor lock-in.

Open the workbench See what's inside a SKILL
  • Plain Markdown
  • YAML manifest
  • Diff-friendly
SKILL: weekly-status-recap files: 6 · examples: 3 · refs: 2 
// SKILL.md - frontmatter
---
name: weekly-status-recap
version: 2.1
description: |
  Turns raw meeting notes and PR titles into a
  one-page status update in the team's voice.
inputs:
  - notes: raw meeting notes (markdown)
  - prs:   list of PR titles (one per line)
  - week:  ISO week number
output: ./examples/output-shape.md
references:
  - ./refs/voice-guide.md
  - ./refs/exec-glossary.md
---

// Role
You are the chief of staff for an eng team of 12.
Be terse. Lead with the verb. No adjectives.

// Rules
1. Decisions first, status second, risks third.
2. One sentence per item. No "we are excited to".
3. Quote a number whenever you can.
Plain Markdown Every SKILL is a directory of .md files plus a YAML manifest. Open it in any editor. No proprietary container.
Versioned & diffable Drop the directory into Git. Track every edit to instructions, examples, and references like you track source code.
No runtime dependency Build it in Prompt Organizer, run it in Claude Code, Cursor, Copilot CLI, Gemini CLI, Codex CLI, or an AGENTS.md repo. The format is portable.

What's inside a SKILL

Four parts. Each one earns its keep.

A SKILL package is not a longer prompt. It's a structured contract. The four parts below are what separate a workflow that runs the same way every Monday from one that drifts every time you re-open the tab. Each lives in its own file, so each can be edited, diffed, and reviewed independently.

weekly-status-recap/ the four files that matter 
// 1. SKILL.md
The role, the rules, the inputs, the output target.
This is what the model reads first.

// 2. examples/
output-shape.md       // the literal format
good-example-w14.md   // from a real Monday
edge-case-no-prs.md   // the empty-input case

// 3. refs/
voice-guide.md        // "lead with the verb"
exec-glossary.md      // what "shipped" means here

// 4. manifest.yaml
Inputs, outputs, version, license. Machine-readable.
The runtime uses this. You usually don't edit it.
  1. Instructions - the role and the rules.Not "be helpful". A specific role, with constraints. "You are the chief of staff. Be terse. Lead with the verb. No adjectives." Constraints are what make outputs predictable.
  2. Examples - input/output pairs from real work.The model copies the shape of the example more reliably than it follows abstract instructions. Three real examples - one normal, one edge case, one failure mode you've already debugged.
  3. References - the context it should pull from.Style guides, glossaries, exec vocabulary. The model doesn't have to guess what "shipped" means at your company - the reference says so. Refs are versioned files, not paragraphs glued into the prompt.
  4. Output spec - the exact shape of the result.Markdown? JSON? A specific table? One paragraph of 60 words or less? The spec is what makes the SKILL composable - the next step in your pipeline knows what to expect.

Three real SKILLs. Three repeat jobs.

SKILLs earn their cost the third time you run them. These are the three patterns we see most often - all of them started as a prompt someone kept retyping, and turned into a packaged workflow once it had a shape worth keeping.

  1. code-review-security-pass (Engineering).Inputs: a diff and a language. Output: an itemized list of auth, escape, and injection concerns with line refs. Refs: your internal security checklist. The same SKILL drops into Claude Code as an agent and into ChatGPT Skills for non-CLI reviewers.
  2. weekly-status-recap (Operations).Inputs: meeting notes plus PR titles. Output: a one-page exec update in the team's voice. Refs: voice guide, exec glossary. Run it every Monday. The output lands in the same shape, so the recipient's eyes know where to look.
  3. doc-rewrite-clarity (Editorial).Inputs: a draft and an audience. Output: a rewritten version plus a three-line diff explaining what changed and why. Refs: your style guide. Used by anyone writing for buyers - the SKILL enforces the house voice without a human editor in the loop.
~/skills/weekly-status-recap/ Pro export 
// Exported SKILL package - portable Markdown
weekly-status-recap/
├── SKILL.md              // role, rules, inputs
├── manifest.yaml         // version: 2.1
├── README.md             // human-readable summary
├── examples/
│   ├── output-shape.md   // the target format
│   ├── good-example-w14.md
│   └── edge-case-no-prs.md
├── refs/
│   ├── voice-guide.md    // "lead with the verb"
│   └── exec-glossary.md
└── .promptorganizer/
    └── source.json       // round-trip metadata

// 11 files · 14 kB · plain text
// Drop into: Claude Code, Cursor, Copilot CLI,
// Gemini CLI, Codex CLI, or any AGENTS.md repo.

11 files · 14 kB · belongs to you · no runtime dependency

When a prompt becomes a workflow, ship it.

SKILL packaging is Pro. One purchase, $39, yours forever. You get ZIP bundles, Markdown exports, SKILL packages, and every future Pro format - all built into the same local-first app. No subscription, no account, no server round-trip.

Unlock Pro exports - $39 once Compare plans

Build it once. Run it anywhere.

SKILLs are Markdown. They diff. They version-control. They survive the next vendor pivot and the one after that. Build them in Prompt Organizer, export them when they're ready, drop them into whatever agent runtime your team uses this quarter. The workflow is yours.

Open the workbench See the prompt manager