00-onboard

Onboard

State-aware onboarding hub for the current project. Silently inspects the project, the AIDD setup, and the installed AIDD plugins, then opens a briefing and a menu: understand the project, set up or refresh the memory bank, or continue the AIDD development journey. Loops until the user stops.

Available actions

#	Action	Role	Input
01	detect-state	Silently probe the project, the AIDD setup, and the installed AIDD surface	project root
02	recommend-next	Render the briefing and the hub menu, route the pick to one concrete action	internal state from 01
03	execute-or-handoff	Carry out the choice: briefing, run, explain, handoff, swap, or stop	choice from 02

Default flow

01 -> 02 -> 03 -> back to 01 after each action, until the user picks Stop. Run each action's ## Test before moving to the next.

Transversal rules

• Hub, not a track. Onboard shows a briefing and a menu of project actions; it never forces a single next step.
• Silent detection. 01-detect-state prints nothing - no state: snapshot, no signal dump, no Analysis: label ever reaches the user. The first visible output is the briefing header.
• Briefing first. Every pass opens with a clean three-line briefing header: project, AIDD setup, standing.
• Categories, never hard-coded foreign skills. Onboard recommends by category - a function such as "technical planning" or "shipping" - and resolves it at runtime against the skills actually installed. It never names a skill or plugin id that is not installed.
• Honest gaps. A category that resolves to no installed skill is reported as a gap, described by function. Onboard never invents a skill to fill it.
• sdlc_phase is a hint, never a verdict. Onboard shows where the project seems to stand but always lets the user pick the SDLC leg. It never assumes a phase is finished or unfinished.
• Numbered choices only. The user replies with a digit. Free-text replies re-render the same menu with a one-line digit reminder.
• Never assume the user will run a skill in this conversation. For a resolved skill always offer: run-in-session, explain, hand off to a new session, swap, stop.
• Always loop back to 01-detect-state after an action runs or is handed off. The project state changes; re-detect before re-rendering.
• Anti-sycophancy. If the user picks an option that conflicts with the detected state, challenge once before complying.
• Wait for an explicit user response between every menu. Do not auto-advance.

Assets

• @assets/state-matrix.md - the hub menu, the SDLC sub-menu, the AIDD journey backbone, and the category resolution rules

References

• The AI tool's native plugin and skill discovery - the runtime source of truth for which categories onboard can resolve to a real skill