Skills (Cerebellum)

49 bundled skills. Unlimited auto-generated skills via Basal Ganglia. Fully editable.

What are skills

Skills are structured knowledge documents that the agent loads before running. Each skill contains a description, tags, usage guidance, and a body of text that gives the agent context on how to approach a class of task.

Skills have a weight (0–1) that controls injection priority, a brain_region that determines how they're categorized, and a when_to_use field that helps the agent decide when to activate them.

Skills don't execute code — they inform the agent's reasoning before tool calls happen.

Bundled skills

49 skills ship with asyncat out of the box:

Planning

`plan`	Structured goal decomposition and task planning
`onboarding-new-codebase`	First-run exploration of an unfamiliar codebase

Workflow

`architecture-review`	Review system design and surface architectural concerns
`code-review`	Review changed code for quality, reuse, and efficiency
`debugging`	Systematic root-cause diagnosis and fix
`error-recovery`	Handle failures, retries, and unexpected states
`refactoring`	Restructure code without changing observable behavior

Development

`agentic-coding`	Write, test, and iterate on code autonomously
`react-patterns`	React component patterns, hooks, and state management
`nextjs`	Next.js routing, data fetching, and deployment patterns
`docker`	Containerize applications and manage Docker workflows
`ci-cd-pipeline`	Set up and manage CI/CD workflows

Execution

`deployment`	Deploy apps to cloud, VMs, or container platforms
`database-migrations`	Write and run safe database schema migrations
`testing`	Write and run tests — unit, integration, e2e
`tdd`	Test-driven development: red → green → refactor
`cron-jobs`	Set up scheduled and recurring tasks

Analysis

`codebase-metrics`	Measure lines, complexity, and test coverage
`data-engineering`	Process, transform, and analyze datasets
`log-analysis`	Extract signal from logs and error streams
`performance`	Profile and optimize slow code paths
`financial-analysis`	Analyze financial data, metrics, and models

Documentation

`documentation`	Write clear technical docs, READMEs, and guides
`email-drafting`	Compose professional email and outreach
`report-writing`	Structure and write formal reports and summaries
`marketing-content`	Write copy, landing pages, and product descriptions
`document-generation`	Generate formatted documents — PDFs, Word, HTML

Specialized

`accessibility-audit`	Audit UI for WCAG compliance and accessibility issues
`api-design`	Design REST and GraphQL APIs with clear contracts
`customer-support`	Draft support replies and triage issues
`incident-response`	Structured runbook for production incidents
`security`	Security review — OWASP, secrets, injection, auth
`sql-queries`	Write optimized SQL for complex data retrieval

Auto-generated skills

The Basal Ganglia engine creates additional skills automatically as you work. These appear in the Skills browser with origin: basal-ganglia and a default weight of 0.8.

Common examples include test-runner, dependency-audit, local-rag, and api-endpoint-explorer — patterns the agent notices and encodes without prompting.

See Self-improvement for how the Basal Ganglia works.

Custom skills

You can create, edit, and delete skills in the Skills browser (/skills) in the web UI. Custom skills support the same schema as bundled skills:

name: my-deploy-flow
description: Our team's standard deployment checklist
brain_region: cerebellum
weight: 0.9
tags: [deployment, production]
when_to_use: When deploying to production or staging
created_by: user

---
Always run migrations before deploying.
Check the rollback procedure before triggering the deploy.
Notify the #deploys channel with the PR link.

Alternatively, you can add custom skill Markdown files directly to the ~/.asyncat/skills/ directory on your filesystem.