aios-explore

AI-OS Explore

Your job is to show the user a clean, complete menu of every AI-OS skill available to them right now — not a generic list, but a personalized one that reflects what is actually installed in their current session.

How to respond

Do not ask questions. Do not narrate what you're about to do. Output one clean block, styled like a well-designed help page. Keep it scannable.

Structure

Open with one short line that orients the user. Then four sections, in this order:

1. Every day — the skills most people reach for
2. Deep work — strategic and reflective skills
3. Autonomous — skills that run on their own or prepare outputs while the user is away
4. System — meta-skills for managing AI-OS itself

Close with one line pointing to the brain folder.

Content

Only list skills that are actually installed. Check .claude/aios.config.json for installed_skills, and skip anything the user doesn't have (e.g. meeting-prep is absent if they never connected GCal). If no config exists, list the canonical set below.

Use this canonical grouping and copy:

Every day

• /memory-search — pull context from the brain before you start. Use when you want Claude to know the backstory.
• /reflect — an honest, goal-by-goal progress check. Use when you feel scattered or at the end of a week.
• /foresight — 3–7 ranked priorities for the window ahead. Use when planning a new week or recovering focus.
• /project-status — decision-grade read on one project: cadence, state, next action. Use before status meetings or when something feels stuck.
• /decision-check — search prior decisions before making a new one. Use to prevent contradictions.
• /relationship-check (if installed) — last contact, open commitments, next move on one person.
• /meeting-prep (if installed) — one-screen brief for a calendar event.

Deep work

• /brain-dump-content (if installed) — turn raw thoughts into structured content.
• /content-interview (if installed) — Claude asks, you answer, a draft gets made.

Autonomous

• /nightly-consolidation — process short-term memory, extract patterns, improve skills. Runs nightly if you install the scheduler.
• /nightly-goal-pursuit — advance one long-term goal overnight. Writes a morning-delivery file to act on at wake-up.
• /behavioral-learning — extract behavioral patterns from archived transcripts. The brain watching how you actually work.
• /auto-outreach-queue — draft personalized follow-ups for people in your brain. Nothing is sent without your approval.
• /signal-calibration — tune the signal detector that routes context into sessions.

System

• /aios-start — already run. Refuses to re-run (won't overwrite your brain).
• /aios-update (if available) — non-destructive update of skills and hooks.
• /aios-explore — this menu.
• /forge-skill <mcp> — auto-generate intent-wrapped skills from a connected MCP. Run this whenever you connect a new MCP (Gmail, Notion, Linear, etc.) and the brain will propose 3–5 skills tuned to how you work.

Output rules

• One line per skill. Slash command first, em-dash, then the purpose.
• Do not explain the brain model.
• Do not list skills the user doesn't have.
• Do not list integrations that are not connected.
• Do not add a "Getting started" section — that's what /aios-start was for.
• Keep the whole output under ~35 lines.

Close with

One line:

Everything lives in this folder. Edit it by hand or ask Claude to edit it for you — the brain updates itself either way.

Stop there. Do not ask what to do next.

Tick setup-progress

After rendering the menu, silently tick the first-use checkbox so /aios-help status reflects that the user has now seen the skill catalogue:

bash $CLAUDE_PROJECT_DIR/system/scripts/tick-progress.sh "First-use validation" "aios-explore" "aios-explore"

Idempotent — safe to re-run on every invocation. If system/setup-progress.md doesn't exist (legacy install), the script exits silently.