The App Store
A catalogue of skills for AI agents — published by Activated Cloud and by independent developers, installable into Activated Cloud and into other agents.
The App Store at activated.cloud/skills is where agent skills are published and found. Activated Cloud publishes hundreds of skills there; independent developers publish their own. Every skill has one address, publisher/skill — for example activated/notion — and a page that shows what the package actually contains.
What a skill page shows
- Description and documentation — the documentation is the skill's own SKILL.md, the exact text an agent reads when it loads the skill.
- Compatibility — which kinds of agent it works in, and how well (see below).
- Versions — the current version and the history, with the publisher's changelog.
- Requirements and security — the permissions the package appears to need (network, files, shell, secrets), its dependencies, and the result of the security scan.
- Reviews — ratings from people who installed it, with replies from the publisher.
- Publisher — who made it, their verification level, and their other skills.
Finding skills
Search matches names, descriptions and tags. Filters narrow the results to Free, Paid, Open source or Verified skills, to skills that work with a particular kind of agent, and to a category. The store home groups the catalogue into Featured, Trending, Most installed, Top rated, New, Free, Paid, Open source and Verified shelves, plus one shelf per category.
Compatibility
Compatibility is worked out from what the package contains — a SKILL.md, scripts, tool definitions — never guessed. A publisher can mark a target as partial or unsupported, but cannot claim a target the analysis found nothing for.
| Status | Meaning |
|---|---|
| Native | The skill directly supports this kind of agent. |
| Adapted | The store knows how to translate it for this target, and the install plan says how. |
| Compatible | Installs without significant changes. |
| Partial | Some of it works there. |
| Not supported | It cannot run in this kind of agent. |
| Unknown | Nothing in the package let us tell. Not a no — just unverified. |
Security
Every submission is scanned for network access, file access, shell commands, secrets and dependencies, and the findings are listed on the page. The scan is automatic and shallow by design; it tells you what to look at, not that the skill is safe.
| Label | Meaning |
|---|---|
| Scanned | The automatic scan ran and found nothing alarming. |
| Warning | The scan found patterns worth reading before you install — piping downloads into a shell, for instance. |
| Verified | A person at Activated Cloud read the package. This is set by hand, never by the scanner. |
| Unverified / Unknown | Not reviewed, or not scannable. |
Read the requirements
A skill that needs the network or the shell is not a problem in itself — most useful ones do. Just check that what it asks for matches what it says it does.
Reviews
Anyone can leave one review per skill: a name, a rating, a few lines, and optionally the version and the kind of agent it was used in. Your email is asked for so that one person can only review a skill once; it is stored as a hash and never shown. A small arithmetic check keeps bots out. Publishers can reply under a review.