Aevum Cadence AI

Skills reference

What a Cadence assistant can do out of the box, plus how to extend it. This page is a catalog. The live inventory on a hatched assistant is assistant skills list. Mechanics: Skills and tools. Authoring: Your first skill.

Most service skills need a one-time OAuth or API-key connect. Sends, purchases, and host access ask first unless you write a trust rule.

Each skill, when used well, answers:

Communication

Gmail
Archive, label, draft, send, unsubscribe, filters. OAuth. Sends need approval.
Email (Agent Mail)
Give the assistant its own mailbox via a mail provider you configure (not an Aevum-hosted inbox).
Messaging
Read, search, send, and manage messages across connected platforms.
Phone and SMS
Twilio voice and SMS after you store Twilio credentials. Inbound voice can reject. A2P to US cells needs carrier registration when you leave trial.
Slack
Scan channels, summarize threads, send, reactions. User integration and Slack bot are different connects.
Contacts
People, channels, and who may talk to the assistant.
Notifications
Route alerts across the channels you have connected.

Calendar and work

Google Calendar
View the schedule, create events, check availability. OAuth.
Tasks
Reusable templates and a prioritized queue.
Schedule
One-shot and recurring jobs (cron, RRULE, or a fire-at time). See Scheduling.
Followups
Track outbound messages still waiting on a reply.
Playbooks
Trigger-action rules for inbound messages.
Start the Day
A briefing from weather, calendar, news, and tasks you have actually connected.
Document
Long-form drafts: posts, articles, reports.

Web, host, and automation

Browser
Navigate, click, type, extract, screenshot via assistant browser. See Features.
Host computer access
On Linux, host_file_* and host_bash after you grant host access. Not a silent full-disk grant on hatch.
Computer Use / Screen Watch
macOS GUI click, type, and OCR. Not the Linux 0.1 path. Linux uses host tools and Browser instead.
Watcher
Poll an external source and notify on change.
Subagent
Spawn background work that runs in parallel.
ACP
Delegate coding to an external agent over the Agent Client Protocol.
App Builder
Build an interactive web app from a description, in the conversation.

Media and services

Image Studio
Generate and edit images when an image model is configured.
Media processing
Ingest video, audio, and image files through a processing pipeline.
Transcribe
Audio and video to text with Whisper.
Weather
Current conditions and multi-day forecasts.
Amazon
Search and shop after you connect the account. Purchases need approval.
DoorDash
Order food or groceries after you connect. Purchases need approval.
ChatGPT import
Import conversation history from a ChatGPT export into this assistant.

System and setup

Skill management
Create, inspect, and delete custom skills in the workspace.
Skills catalog
Discover and install first-party and community skills (including skills.sh).
Cadence self-knowledge
Answer "can you do X?" from the CLI, these docs, and source. Not from guesswork.
LLM provider setup
Bring your own keys (Hive, xAI, or another OpenAI-compatible endpoint). Cadence does not bill model usage in 0.1.
OAuth / MCP / channel setup
Dedicated setup skills for Slack, Telegram, Twilio, GitHub, Linear, Notion, and MCP servers. "Is X connected?" has two answers: catalog vs live status. Use assistant oauth status and assistant channels list.

Related: What is Cadence, Key concepts, FAQ.