Keep instructions and references together.
Publish the method with the text files it depends on: approved wording, checklists or examples. Your team can read the available skill in Wardyn and retrieve it through a supported assistant.
what wardyn adds
Your company’s instructions. Ready to share and maintain.
A skill is a set of AI instructions and supporting references for a particular task.
A shared folder can hold the files. Wardyn adds a place to publish a chosen version, make it available to the right people, and review the next change.
Choose which published skills each managed MCP connection offers. That connection lets a supported assistant retrieve skills for people you authorize.
Published company skills
Team members’ assistants
one skill in use
Turn your best AI instructions into a skill your team can use.
A proposal that starts with your company’s way of working.
Priya has notes from a client meeting. Her company’s proposal skill supplies the instructions and approved service wording she wants the draft to use.
Her company’s skill contains
“Use the approved wording for discovery. Separate agreed work from open questions. Ask before adding a delivery date or price.”
“Discovery includes one workshop and a written project brief. Design and development are scoped separately.”
Priya brings
“The client needs a website redesign. Start with discovery. The launch date is not agreed.”
“Use our proposal-writing skill to draft the next steps from these meeting notes. Ask me for anything missing.”
Illustrative passage for Priya to review
“We’ll begin with a discovery workshop and a written project brief. Design and development will be scoped separately.”
Question for Priya
“What launch date should we plan around?”
Illustrative company skill and sample passage · your company supplies the content. This is not an included template, a product screenshot or a recorded AI result.
Wardyn supplies the published instructions and references. The assistant drafts; Priya checks the result before sending it.
Then your service changes.
A teammate reports that discovery now includes two workshops. The skill’s owner prepares an updated package. An administrator compares the reference change and publishes the new version.
- Before:
- “Discovery includes one workshop and a written project brief.”
- Proposed:
- “Discovery includes two workshops and a written project brief.”
Connections following the latest publication retrieve the new version on future requests. Existing conversations are not rewritten.
features
What you can do with Wardyn.
Choose who can use which skills.
Select the skills offered through a managed MCP connection and control who can use it. Give sales its proposal skill and support its reply-writing skill, with shared skills available to both where appropriate.
Review the changes before publishing.
Compare versions to see which instructions and supporting files were added, changed or removed. An administrator chooses the version to publish.
Turn feedback into a reviewed change.
Colleagues can report what worked or what needs changing. Authorized reviewers can ask AI from OpenAI to propose a correction, review it and create a draft. Publication remains a separate administrator action.
See which skills your team requests.
Administrators can review recorded skill requests over 7 or 30 days, see recent requests and spot request errors—without recording chat conversations in the usage log. Records can be incomplete and do not measure completed work.
Keep an owner responsible for each skill.
Give each skill a named owner. Administrators can transfer ownership when responsibilities change, so there is someone to keep the instructions useful.
access and package checks
Share skills with clear controls.
Check uploads.
Skill packages are checked for malware, unsafe archives and recognizable secrets, such as access keys.
Release a chosen version.
Publication releases a fixed version. A live connection and the person’s access determine whether the assistant can retrieve it.
End future access.
Revoke access when it should end. Wardyn refuses the next unauthorized request for the affected content. Material already delivered may remain in conversations or saved copies.
getting started
Start with one skill. Invite your team when you’re ready.
You can add and try your first skill on your own.
1 — Create your workspace.
Sign up, verify your email and create your company workspace.
2 — Add and try a skill.
Upload its instructions and supported text references, review the package and publish it. Select the skill for your own managed MCP connection, connect a supported assistant and try it.
3 — Share when you’re ready.
Invite teammates and give them access to the connections they need. Each person connects their assistant.
the practical details
Questions before you start.
Do we need Wardyn if our assistant already lets us share instructions?
Your assistant’s sharing features may be enough for your team. Consider Wardyn when you want a place to maintain company skills, choose what to publish, manage access through supported assistants and review feedback and recorded requests together.
Do you provide the skills?
You bring your company’s instructions and supported references. Wardyn helps you publish, share and maintain them. The proposal above is an example of content your company could provide.
What is an MCP connection?
MCP, short for Model Context Protocol, is a standard for connecting AI assistants to tools and information. Wardyn uses it to make your published company skills available in supported assistants, with access your company controls.
Where can we use the skills?
In a supported AI assistant, through a managed MCP connection. Your assistant’s plan and company settings can affect setup. Check the setup guidance before connecting.
What happens when I publish an update?
Connections following the latest publication use the new version on future requests. Connections set to an older version stop offering that skill until an administrator updates them. Existing conversations are not rewritten.
Does AI change our skills automatically?
No. Authorized reviewers can request an AI-drafted correction. Skill instructions, supporting text references and selected feedback notes go to OpenAI to propose changes. People review the proposal, create a draft and publish separately.
Start with one skill your team would use again.
Create your workspaceStart with your workspace. Invite your team when you’re ready.