Agent Skills

Skills are prescribed commands that a user or AI agent can invoke to query Autographical. They act as thin launchers — they fetch activity data, signal intent, and let the AI lead the conversation from there.

Skills

/whats-up — What’s been happening

Fetches recent activity and brings context into the session.

When to use: Starting a new session, picking up where you left off, or mid-session to refresh with the latest activity.

What it does:

1. Calls recent_activity to fetch recent sessions and top resources
2. Optionally calls search if the user provides arguments to narrow focus
3. Can drill deeper into specific sessions or resources via get_session and get_resource
4. Synthesizes the data into a summary — active threads, momentum, attention, recency

The activity stream includes human activity (browsing, reading, saving) and AI activity (Claude Code sessions, ChatGPT conversations, and other tools).

/whats-next — What to do next

Suggests what to work on next based on recent activity and the current working context.

When to use: At a decision point, between tasks, or when you want direction.

What it does:

1. Injects local context automatically (working directory, git repo, branch, recent commits)
2. Calls recent_activity and optionally search
3. Can drill deeper into relevant paths or resources
4. Leads with an opinionated recommendation — which paths have open loops, what connects to the current repo, what’s been dropped

Agents

Claude Code

Skills are available for Claude Code via a plugin. See Claude Code — Skills.