> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orcapods.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Shared kits
> Publish a working pod, agent, or scheduled run as a link anyone can copy into their own workspace.
A **kit** is a frozen snapshot of something you have already built and tested: a [pod](/concepts/pods), a single [agent](/concepts/profiles), or a scheduled run, together with the [skills](/concepts/skills) it depends on. Publishing one produces a public link. Anyone who opens that link can install their own editable copy in a single action.
A kit is a snapshot, not a live connection. Copying instantiates the stored payload into the copier's workspace and never reads the author's live one, so neither side can see into the other.
Not to be confused with the in-app **Quickstart templates** gallery, which pre-fills the new-agent dialog with starter configurations. Those ship with Orca and are not shareable. A kit is published by a workspace, carries a public link, and can be copied between accounts.
## What a kit can carry
| Asset | What it is | Renameable on copy |
| ------------ | -------------------------------------------------------------------- | --------------------------------------------------------------------------------------- |
| `skill` | A [skill](/concepts/skills) package with its reference files | Yes |
| `profile` | An [agent](/concepts/profiles): runtime, model, system prompt, tools | Yes |
| `pool` | One [pod](/concepts/pods) and its member roles | Yes |
| `automation` | One scheduled run, with its cron expression and timezone | Yes |
| `poolFile` | A file from the pod's shared workspace | Yes, subject to path-safety checks rather than the plain name rule used for other kinds |
A kit carries at most one pod, alongside any number of agents, skills, and scheduled runs. The kit page labels itself by the largest thing it contains: **Scheduled run** if it has an automation, otherwise **Pod**, otherwise **Agent**.
## The share link
Every kit is addressed by a stable public id rather than by slug and author handle. Its canonical link is:
```
https://app.orcapods.ai/kits/
```
The **Copy link** button on the kit page puts this URL on your clipboard. An older link in the shape `/templates?org=&template=` still works: it redirects to the same `/kits/` page, so links published before the switch keep working.
The link works for people who have never heard of Orca. A signed-out visitor sees the same kit card as a signed-in member. If they choose to copy it, Orca stores the handoff for that browser tab, walks them through sign-in and workspace creation, then returns them to the same kit page. The stash expires after 30 minutes, so a half-finished signup does not resurface on a later visit.
The public-id shape is what makes a kit link safe to post outside Orca, such as on Reddit or X: the id never changes even if the author later renames the kit or its label. A signed-out visitor to a kit page also gets a **Contact us** link, so someone with a question before signing up can reach Orca directly.
## Save and copy are different
Bookmarks the snapshot in your workspace's saved list. Nothing is installed and nothing runs. Use it when a kit looks useful but you are not ready to set it up.
Opens the copy plan and installs editable assets in your workspace. From that moment they are ordinary agents, skills and pods that you own, edit and delete like any others.
You cannot save your own kit. Authors already have the original.
## The copy plan
Copy never installs blindly. It first shows a plan with one row per asset, each carrying a status:
| Status | Meaning |
| ------------- | ----------------------------------------------------------------- |
| **new** | You have not copied this asset from this kit before |
| **changed** | You copied it before and the author has since changed it |
| **unchanged** | You copied it before and it is identical to what you already have |
Each row also carries a **target name**, the name the asset will take in your workspace. If that name is already taken, the row is flagged `name taken` and Orca suggests a free one by appending `-copy`, then `-copy-2` and upwards until it finds one. You can type any name you like instead, or untick the row to skip that asset entirely.
**Nothing in your workspace is ever replaced.** A name collision is always resolved by renaming the incoming asset or skipping it, never by overwriting what you already have.
Renames are safe. When you rename a skill, every agent in the same copy that referenced it is rewritten to point at the new name, and the same holds for agents inside a pod and for the pod a scheduled run targets. The kit arrives wired together whatever you call its parts.
Target names are limited to 128 characters and may not contain slashes or line breaks.
## What is never copied
The sanitizer strips these before a snapshot is stored, regardless of what the author selected:
* **Environment variable values.** Variable names survive, so you can see what the agent expects. The values are blanked.
* **Literal MCP header values.** A header whose value is a `secret://` or `${VAR}` reference survives as a reference. A header holding a literal value is removed.
* **Row identifiers and computed fields**, so a copy never carries the author's internal ids.
[Connected apps](/concepts/connected-apps) are never part of a kit either. If a kit's agents use GitHub, Slack or any other connection, the person copying it connects their own account afterwards. That is deliberate: a kit hands over a method, never a credential.
Pod workspace files are excluded by default when publishing, because that is where site-specific and personal data usually lives. Authors can opt individual files in.
## Updates
A kit has a version, which increases when the author republishes. If you have copied a kit before and the author has since changed it, the kit page shows **Update available**.
Copying again is safe and incremental. The plan marks assets you already hold as `unchanged` so you can untick them, and shows only genuinely `changed` assets as worth taking. Your renames are remembered, so an update lands on the names you chose the first time rather than creating a second set.
## Publishing your own
From a pod, an agent, or a scheduled run, choose **Copy as Orca kit**. Orca computes a manifest of everything the snapshot would contain and shows it for review before anything is published:
Skills can be ticked or unticked. Pod workspace files are unticked by default. A locked section lists what the sanitizer removes no matter what you choose.
A label and a slug. The slug becomes part of the public link, so it is lowercase kebab-case, 2 to 64 characters.
The description shown on the kit page. Plain lines become paragraphs and lines beginning with `- ` become bullets. Up to 4000 characters, editable afterwards from **Edit description** on the kit page.
The dialog hands back the shareable link.
As the author you see a small activity tally on your kit page: how many other workspaces saved, copied, and removed it. Your own actions are excluded, so testing your own link does not inflate the numbers. Visitors never see these.
### Republishing someone else's snapshot
A kit you have saved or copied can be republished under your own workspace with a new slug and label. That creates a separate kit with its own link, owned by you. The original is untouched and keeps its own link.
### Withdrawing a kit
Deleting a kit withdraws its public link. Copies that other workspaces already made keep working, and saved snapshots stay readable. Deleting does not reach into anyone else's workspace.
## Try one
A complete walkthrough of one kit, from opening the link to reading the first pull request in your inbox.
The source of the kits Orca publishes: agent documents, skills and the prompts that point them at your product.