List the SkeletIQ projects this token can see, optionally filtered by name. Use this first to turn a project a person named in conversation into an id. If more than one project matches, ask which — do not pick one.

Read a SkeletIQ architecture. Start with mode "brief" to orient — it returns a fenced block to write into AGENTS.md — then "build_order" for the sequence and "component" for each piece as you build it. Every answer states which version it came from and whether that version is a release.

Design a system architecture from a prompt. This spends the account holder's credits and takes several minutes, so do not call it speculatively — if a design already exists, read it with get_design instead. Which model runs it is the account holder's stored setting; you do not choose it. Describe the system, what it must do, and the scale and constraints it runs under: whatever the prompt leaves out is assumed, and a prompt that names no system at all comes back as clarification_required with the questions to answer.

Check a background generation started with wait: false. When status is "completed", read the result with get_design.

Check a design against SkeletIQ's architecture rules and get scored findings back. Deterministic and free — no model call, no credits, nothing stored. Useful on a design you drafted yourself before committing to it.

Tell SkeletIQ which components you have built and get back what is missing, what is only half done, and what exists in the code but not in the design. Declare covers when this repository implements only part of the design. This changes nothing in SkeletIQ — an agent never writes back to a design.