bcskills.ai
Authors

Author guidelines

What makes a good skill, what a package may contain, and what happens after you submit.

What makes a good skill

  • One job, done well. A skill should teach an agent one clear task or area, not everything about Business Central.
  • A precise description. Agents decide whether to load your skill from its description. Say what it does and when to use it.
  • Real examples and the exact steps, field names and page names an agent needs.
  • Honest scope. State the Business Central versions and platforms it was tested with.
  • Screenshots with alt text when the skill has a visible result.

The package

Upload a .zip (up to 4 MB) or a single SKILL.md. The package needs a SKILL.md with frontmatter:

---
name: Post Sales Invoices from PDF
description: Reads a customer PO PDF, finds items and prices in Business Central, and drafts a sales invoice. Use when a user supplies a purchase order PDF.
---

# Post Sales Invoices from PDF
Step-by-step instructions for the agent…

Both name and description are required. Your listing summary (160 characters at most) becomes the meta description for search engines.

Allowed files

  • Markdown, text, JSON, YAML, CSV, AL source, templates, images and other plain-text resources are fine.
  • No binaries. Executables, libraries, installers and archives are rejected (for example .exe, .dll, .so, .jar, .wasm).
  • Scripts are flagged. Shell, PowerShell, Python, JavaScript and similar files are allowed, but the listing shows a “Contains scripts” badge and reviewers read them closely. Explain what each script does.
  • Never include secrets, API keys, customer data or tenant identifiers.

Licenses

  • MIT, free to use, change and share.
  • PolyForm Shield 1.0.0, free to use, but not to build a competing product from it.

You must own what you upload or have the right to share it. Skills copied from other people without permission are removed.

How review works

  1. You submit a version. It is immutable and gets its own checksum.
  2. Automated review runs first: package checks, a secrets scan and an AI summary of what the skill does and any risks.
  3. A moderator then approves it, asks for changes or rejects it, with a note to you.
  4. Approved versions go live. Every version is reviewed, including updates from established authors.

Ready?

Publish a skill. You will sign in with an emailed code, no password needed.