Skill Generator From raw knowledge to ZIP-ready skill bundles

Generate portable SKILL.md bundles with source-backed structure, scripts, and safety gates.

Skill Generator turns URLs, documents, pasted notes, and product intent into agent-ready bundles for Codex, Claude, OpenAI Skills, and OpenClaw. It builds the folder structure, drafts the core instructions, and flags risky executable content before you ship it.

SKILL.md + ZIP portable output
URL / DOC / Paste source ingestion
Clean / Warning / Danger risk verdicts
Bundle

Skill-ready by default

Creates `SKILL.md`, `references/`, optional `scripts/`, optional `agents/openai.yaml`, a test prompt set, and a ZIP you can hand off immediately.

Research

Official-source aware

Pull from a URL directly, ask for source research, or upload a document so the bundle is grounded in concrete material instead of generic filler.

Safety

Warnings before runtime

Heuristic review runs automatically. Paid plans can ask the model to classify scripts and instruction flow as clean, warning, or danger.

Why This Exists

The current skills ecosystem expects folders, not just one text blob.

Skill Generator is designed around the way agent skills actually work: progressive disclosure, optional scripts and references, platform-specific add-ons, and security review for anything executable. The output is built to be download-ready and database-ready.

Turn expertise into structured bundles

Use the walkthrough to capture trigger phrases, non-goals, runtime constraints, and validation steps instead of hoping the model guesses them.

Bring in real source material

Fetch a URL, upload a document, paste SOPs, or research official documentation so the generated bundle has real grounding and better references.

Review scripts before anyone runs them

Executable content is treated as a risk surface. The tool calls that out visually and supports deeper paid-plan scanning before submission.

Step 1

Define the skill

Sign in to generate bundles and submit them.

Target platforms

These shape the generated metadata and add-ons.

Step 2

Add source material

Pull in the knowledge the model should turn into the bundle.

Source Pack

Attached inputs

0 sources
No sources added yet. You can still generate from the brief alone, but the output will be less grounded.
Step 3

Runtime and governance

Execution choices

These become hard constraints in the generated instructions.

Generation packet

What the model will receive

Structured and versionable

This creates one bundle draft, runs a heuristic risk pass, and prepares a ZIP export. AI risk scanning is a separate paid-plan action.