Domain: ai-agents Depends on: — Depended by: — Implements: 📦 Installable agent skills — smart applications any agent can drop into a projec... Related: —
Smart applications any agent can install. Clone, configure, run. No fleet membership required.
Every skill in this registry is a self-contained application that installs by cloning, gets better with feedback, and works without central authority. No monorepo. No workspace. No API gateways.
An agent has a job to do — monitoring a service, spreading information across the fleet, generating dream content for training. Instead of writing that capability from scratch, the agent clones a skill.
git clone https://github.com/SuperInstance/agent-skills.git ./skills
The skill comes with its own README, its own tests, its own configuration. The agent customizes it, runs it, and — if they choose — contributes improvements back.
| Skill | What It Does |
|---|---|
| fleet-murmur | Fleet-wide murmur propagation — agents spread signals with controlled entropy |
| fleet-spread | Information dissemination — firehose or drip, topic-filtered |
| plato-client | PLATO tile client — read, write, subscribe to room updates |
| mermaid-engine | Generate mermaid diagrams from structured data |
| dream-engine | Pseudo-random content generation for synthetic training |
| swarm-monitor | Fleet health monitoring and alerting |
Each skill has a feedback/ directory. When an agent customizes a skill and improves it, the improvement can be contributed back. The skill doesn't have a single maintainer — it has a fleet of users who make it better through use.
- agent-skills — installable intelligence (this)
- agent-forge — the framework that runs skills
- bootstrap-spark — onboarding that discovers installed skills
- baton-skill — handoff that preserves skill state
MIT