setup · diff
git:20260707.2a2e857 to git:20260710.d9f833e
44 added, 86 removed. Audit A to A.
---
name: setup
- description: First-run setup — configure aigent-OS to know who you are and how you work
+ description: Configure or reconfigure aigent-OS identity, priorities, authority, projects, people, and decision logic
trigger: /setup
---
- # aigent-OS Setup
-
- Run this once after cloning the repo. aigent-OS will ask you a series of questions and configure the system documents and vault to match your specific context.
-
- ## Detection
-
- If `system/00_identity.md` still contains the placeholder text "Replace this section with context about yourself", the AIgent should suggest running `/setup` automatically on first `/open`.
-
- ## The Interview
-
- Walk through these sections one at a time. Ask naturally — this is a conversation, not a form. Adapt follow-up questions based on what the user shares. Don't ask all questions at once.
+ # aigent-OS setup and reconfiguration
- ### Section 1 — Identity (→ writes to `system/00_identity.md`)
+ `/start` owns the day-one onboarding arc. `/setup` is the deeper configuration flow and may also be used later to revise one section.
- **Ask:**
- - "What's your name and what do you do? Give me the quick version — role, businesses, responsibilities."
- - "How do you like to work with AI? Direct and fast? Collaborative? Do you want me to push back on bad ideas or just execute?"
- - "What's your risk tolerance? Are you cautious and methodical, or do you move fast and fix things later?"
+ ## Entry behavior
- **From the answers, fill in:**
- - The "Your Principal" section of `system/00_identity.md`
- - Optimization targets (what matters most to this person)
- - Communication style preferences
+ Read `.aigent/state.json` first.
- ### Section 2 — Priorities (→ writes to `vault/memory/ACTIVE_PRIORITIES.md`)
+ - If state is missing or not `ready`, route through `/start` unless the operator explicitly requested advanced setup.
+ - If state is `ready`, ask which configuration area needs revision rather than replaying the entire interview.
+ - Never detect setup state by searching for placeholder prose in a Markdown file.
- **Ask:**
- - "What are you working on right now? Give me your top 2-3 priorities."
- - "Which of these is the most urgent — the one that needs progress this week?"
- - "Are you in growth mode, or are you stabilizing / fixing things right now?"
+ Before writing, set `status: setup-in-progress` while preserving any prior `completedAt`. On interruption, the next `/start` or `/setup` should resume from the last durably completed section.
- **From the answers, fill in:**
- - Tier 1, 2, 3 priorities
- - Operating mode (Stabilization / Build / Expansion)
- - Any current blockers
+ ## Interview
- ### Section 3 — Authority Boundaries (→ writes to `system/12_authority_matrix.md`)
+ Ask one question at a time and write each completed section immediately.
- **Ask:**
- - "What should I be able to handle without asking you? Think: routine stuff, research, organizing."
- - "What should I recommend but wait for your approval on? Think: strategic moves, external communications, spending."
- - "What should I never touch? Hard lines — things you always want to decide yourself."
- - "Is there a dollar amount where spending decisions escalate to you? Like, anything over $X needs your sign-off?"
+ ### 1. Identity
- **From the answers, customize:**
- - Level 1 (Autonomous) list with their specific examples
- - Level 2 (Recommend & Confirm) list
- - Level 3 (Human Only) list
- - Dollar threshold for financial escalation
+ Ask for the operator's role, responsibilities, preferred collaboration style, and risk posture. Update `system/00_identity.md` without replacing framework-owned doctrine.
- ### Section 4 — Decision Logic (→ writes to `system/14_decision_framework.md`)
+ ### 2. Priorities
- **Ask:**
- - "When you're deciding whether to pursue something new, what do you look for? What makes you say yes vs. no?"
- - "How many things can you actively work on before quality drops? 2? 3? 5?"
- - "What's your biggest time trap — the kind of work that feels productive but doesn't actually move the needle?"
+ Ask for the top two or three active priorities, the most urgent outcome, operating mode, and blockers. Update `vault/memory/ACTIVE_PRIORITIES.md`.
- **From the answers, customize:**
- - Pattern filter (what they look for in opportunities)
- - Active bet limit
- - Anti-drift rules specific to their patterns
+ ### 3. Authority boundaries
- ### Section 5 — Projects & People (→ writes to `vault/projects/` and `vault/people/`)
+ Ask what the AI may handle autonomously, what needs confirmation, what is human-only, and whether spending has an escalation threshold. Update `system/12_authority_matrix.md`.
- **Ask:**
- - "Let's set up your active projects. For each one: what is it, what's the current status, and what's the priority?"
- - "Who are the key people I should know about? Partners, team members, key contacts — and what's their role relative to you?"
+ ### 4. Decision logic
- **From the answers, create:**
- - One vault note per project in `vault/projects/` using the project template
- - One vault note per person in `vault/people/` using the person template
- - Update wikilinks in `ACTIVE_PRIORITIES.md` to reference the new project notes
+ Ask what makes the operator accept or reject opportunities, their active-work limit, and their most common time trap. Update `system/14_decision_framework.md`.
- ### Section 6 — Agents (→ writes to `vault/agents/`)
+ ### 5. Projects and people
- **Ask:**
- - "Do you want to set up any specialized agents now? For example: an engineering agent for code, a research agent for investigation, a creative agent for content? Or do you want to start with just me and add agents later?"
+ For each active project, capture purpose, current state, priority, next action, and relevant people. Create or update notes under `vault/projects/` and `vault/people/`, then connect them from active priorities with wikilinks.
- **If yes, for each agent ask:**
- - "What should this agent be called?"
- - "What's its job — what scope does it own?"
- - "Should it run on a fast model (cheap, for simple tasks) or a mid model (smarter, for real work)?"
+ ### 6. Specialist agents
- **Create** vault notes in `vault/agents/` using the agent template.
+ Ask whether any specialized agents are useful now. For each accepted agent, capture name, scope, tools, model tier, escalation boundary, and success criteria. Store the definition in `vault/agents/` with valid `name:` and `tools:` frontmatter so the installer can register it in `.claude/agents/`.
- ## After Setup
+ ## Completion
- Once all sections are complete:
+ After all requested sections are durably written:
- 1. Confirm what was configured: "Here's what I set up: [summary]. Want to adjust anything?"
+ 1. Summarize exactly what changed.
+ 2. Append an idempotent setup entry to `vault/memory/SESSION_LOG.md`.
+ 3. Set `.aigent/state.json` to:
- 2. **Teach the session rhythm** — this is critical. Say something like:
+ ```json
+ {
+ "schemaVersion": 1,
+ "status": "ready",
+ "completedAt": "<ISO-8601 timestamp>"
+ }
+ ```
- > "One last thing — how aigent-OS sessions work:
- >
- > **`/open`** — Run this at the start of every session. I'll load your context, check what's pending, and surface anything that needs attention. Takes a few seconds.
- >
- > **`/close`** — Run this at the end of every session. I'll save what happened, update the vault, and set up the next session so you never lose context.
- >
- > These two commands are the heartbeat of the system. `/open` boots me up with full awareness. `/close` makes sure nothing falls through the cracks. If you skip `/close`, the next session starts blind.
- >
- > Ready to go? Type `/open` to start your first real session."
+ 4. Write `.aigent/first-run-done` for compatibility with older installs.
+ 5. Teach or remind the operator of `/open`, normal work, and `/close`.
- 3. Note in `vault/memory/SESSION_LOG.md`: "Initial setup completed. System configured for [name]."
+ Do not mark the installation ready when required writes failed.
## Rules
- - **One section at a time.** Don't dump all questions at once. Have a conversation.
- - **Adapt to what they share.** If someone gives a detailed answer, skip the follow-ups that were already covered.
- - **Use their language.** If they say "side hustle" don't write "secondary venture." Mirror their vocabulary.
- - **Don't over-ask.** If they seem done with a section, move on. They can always refine later.
- - **Write files as you go.** Don't wait until the end. Update each file after its section is complete so progress isn't lost if the session ends early.
- - **Skip sections if asked.** "I'll do that later" is a valid answer. Move to the next section.
+ - One section at a time.
+ - Adapt to information already supplied.
+ - Mirror the operator's language without inventing details.
+ - Preserve user-authored content and make precise edits.
+ - Sections may be deferred explicitly.
+ - Reconfiguration must not erase unrelated state.