Multi-platform Get Started with ATQ¶
Goal¶
Provide a fast entry point to use atq across devices without forcing people to understand the whole repository from minute one.
This guide is aimed at:
KabehzL0KYAJCG131281- new collaborators
- operators who need to use the system before contributing to the repo
Simple rule¶
If you want to use ATLANTYQA:
- enter through
atq - follow a short route by role or device
- validate through FSM
- leave evidence
If you want to build ATLANTYQA, then go to GET-STARTED.md.
What kind of experience atq tries to provide¶
atq is not meant to force people to memorize internal scripts or guess hidden jumps.
The intended experience is:
- a clear human intention before the command
- one short entry layer that works across platforms
- visible state progression
- one useful action per block
- fast feedback
- local evidence before relying on remote systems
- continuity across mobile, desktop/laptop, and CLI without losing context
Important¶
- the shell can be native
- the repo Python runtime should come from
.venv - if you close the terminal, activate
.venvagain - multi-platform does not mean every device should do everything equally
desktop/laptopis the full route;smartphoneandtabletusually act as reading, validation, and handoff surfaces
Why local-first matters¶
Using atq in local-first mode is not just a technical preference.
It also helps:
- preserve collaborator authorship and intellectual property
- leave reproducible evidence in the repo and
outputs/ - avoid losing useful decisions or prompts inside ephemeral chats
- keep human control over what gets executed, shared, and promoted
10-minute quick route¶
Step 1. Clone the repo¶
bash
git clone https://github.com/atlantyqa-labs/atlantyqa-universe.git
cd atlantyqa-universe
Step 2. Activate the Python environment if you will use repo tooling¶
bash
python3 -m venv .venv
source .venv/bin/activate
python3 -m pip install --upgrade pip
python3 -m pip install -r requirements.txt
Step 3. Enter through atq¶
Canonical examples:
bash
bash scripts/atq device sso describe --audience onboarding
bash scripts/atq human onboarding --mode plan --domain ops --use-case bootstrap-artifacts
bash scripts/atq ops reconcile --audience onboarding --dry-run
Step 4. Validate your node¶
Examples:
bash
bash scripts/atq collab doctor --collaborator-id l0ky
bash scripts/atq collab doctor --collaborator-id ajcg131281
Real device matrix¶
| Device or platform | What is realistically expected | Recommended ATQ entry | Level |
|---|---|---|---|
| Linux desktop/laptop | full local-first operation experience | collab, ops, human, device |
recommended |
| macOS | near-full experience with shell + bash + .venv |
human, ops, device, doctor |
supported |
| Windows with WSL | full experience if the real work happens inside WSL | collab, ops, human, device |
supported-via-wsl |
| Windows without WSL | partial reading and validation; do not assume full parity | portal + GitHub + compatible shell | limited |
| Android / Termux | light bootstrap, operational reading, and quick queries | device, human onboarding, ops reconcile --dry-run |
lightweight |
| Smartphone with GitHub App | review, evidence, decisions, and handoff | portal + issue/PR + mobile CTAs | review-companion |
| Tablet | guided reading, UI/UX review, and handoff | portal + GitHub + remote CLI if needed | companion |
What changes by device¶
Desktop or laptop¶
This is the canonical route when you need to:
- clone the repo
- activate
.venv - run
collab doctor - run validations
- edit files
- prepare strong evidence before a PR
Expected experience:
- open
triad-control-plane - state the goal in one sentence
- run
doctor - choose one short
atqaction - record evidence in
outputs/, PR, or logbook
Smartphone¶
It is not the best surface for changing a lot of code, but it is good for:
- reading the PR
- reviewing portal CTAs and narrative
- validating that the experience does not require zoom or hidden context
- opening issues/PRs from GitHub App
- leaving a short evidence or decision note
Expected experience:
- open the portal or PR
- understand the next step in a few seconds
- confirm whether the real next step must move to laptop/desktop
- leave a short handoff if you do not run local CLI
Tablet¶
It should feel like an intermediate surface:
- better reading than mobile
- less operational power than laptop
- useful for guided review, demos, onboarding, or UI/UX validation
If no real tablet is used in the session, mark not-applicable and do not invent compliance.
Android / Termux¶
It supports a light CLI experience when you want to:
- recall commands
- run quick queries
- validate wrapper narrative and semantics
- prepare a handoff to a stronger node
It should not be presented as an automatic substitute for the full contributor environment.
Short routes by person¶
Kabehz¶
Typical goal: local coordination, triad contract, and route decision.
bash
bash scripts/atq human ops --mode plan --domain governance --use-case triad-sync
What it should feel like:
- intent is clear before execution
- output helps decision-making rather than improvisation
- continuity between agents stays visible
L0KY¶
Typical goal: review, risk, and merge priority.
bash
bash scripts/atq ops reconcile --audience onboarding --dry-run
What it should feel like:
- the command reduces noise
- risk is understandable without opening ten paths at once
- the next review step is clear
AJCG131281¶
Typical goal: first real progress without unnecessary friction.
bash
bash scripts/atq collab doctor --collaborator-id ajcg131281
What it should feel like:
- the first step is obvious
- a block, if present, is explained
- the system invites the user to follow FSM instead of intuition
Recommended usage sequences¶
A. Laptop/desktop first¶
bash
source .venv/bin/activate
bash scripts/atq collab doctor --collaborator-id ajcg131281
bash scripts/atq ops reconcile --audience onboarding --dry-run
Use it when you will execute, validate, and leave strong evidence.
B. Smartphone first¶
- open
docs/index.mdorportal/triad-control-plane/ - open the relevant issue/PR
- decide whether a review is enough or work must continue on laptop
- leave a short note with finding, block, or decision
Use it when reading speed and human decision-making matter most.
C. Cross-device handoff¶
- mobile to read and detect friction
- laptop to run
doctor, validate, and edit - mobile again to verify CTA behavior and final PR state
This is the most realistic multi-platform experience for the current stack.
Minimal FSM to operate¶
graph TD
S0[S0_ENTRY] --> S1[S1_CONTEXT]
S1 --> S2[S2_DOCTOR]
S2 --> S3[S3_ROUTE]
S3 --> S4[S4_ACTION]
S4 --> S5[S5_EVIDENCE]
S5 --> S6[S6_DECISION]
S2 --> F1[F1_BLOCKED]
S4 --> F2[F2_UNCLEAR]
S5 --> F3[F3_EVIDENCE_GAP]
Interpretation¶
S0_ENTRY: enter through a visible routeS1_CONTEXT: state what you are trying to doS2_DOCTOR: validate the environment or collaboratorS3_ROUTE: choose the correct command or portalS4_ACTION: execute one small actionS5_EVIDENCE: record the resultS6_DECISION: decide whether to continue, fix, or block
What good multi-platform UX means in ATQ¶
The experience is good when:
- the same operational language works across mobile, laptop, and CLI
- the user quickly knows whether they are in reading, validation, or execution mode
- portal and commands point to the same contract
- a block is presented as a real block, not as a personal failure
- the handoff between devices keeps context intact
The experience is poor when:
- mobile users have to guess the next step
- desktop users need hidden context
- CLI messages are opaque
- the PR does not make goal, evidence, and pending decision explicit
Minimum evidence expected¶
After using atq on any device, try to leave at least one of these traces:
- a PR or issue comment
- an entry in
docs/internal/bitacora/ - an artifact in
outputs/** - a short summary with
fact,hypothesis, andnext step
Learning by doing¶
Do not try to learn everything at once.
Follow this pattern:
- one command
- one expectation
- one check
- one short result note
That is the core of governed vibe coding: fast exploration, but with validation and shared memory.
Common mistakes to avoid¶
- using global Python instead of
.venvfor repo validations - assuming mobile should replace laptop for contributor work
- opening deep internal runbooks before mastering
atq - executing real changes without
planor without evidence - treating a quick smartphone reading pass as full UI/UX validation
Next layer¶
Once you master the short entry layer:
- architecture:
docs/atq/operating-system-architecture.en.md - operating portal:
docs/portal/triad-control-plane.md - UI/UX validation:
docs/portal/triad-ui-ux-validation-windows.md - contributor setup:
GET-STARTED.md