Architect

Unified runtime invocation

Resolve the plugin root from this loaded file: SKILL.md is at <plugin-root>/skills/<skill-name>/SKILL.md. Invoke only python3 "<plugin-root>/coordinator.py" and send one bounded JSON request on stdin. Before constructing it, read the Coordinator request schema in <plugin-root>/README.md; never invent fields or route/action pairs. The public coordinator re-observes the active host/model, captures artifact provenance, excludes same-family routes, and verifies the co-packaged native manifest. It runs standalone from the installed plugin. Never discover a provider executable or reconstruct a raw command. Frontmatter tier is a routing recommendation, never a coordinator request field. For a review, cross-check, tiebreaker, or fallback over an authored artifact, capture its exact UTF-8 content and observed author model in the optional artifact object even when governance is false; never paste it into the prompt as a provenance substitute.

Architecture consultation

Use the unified coordinator's sealed architecture action for read-only analysis and planning. Grok 4.5 is eligible only through this action (or the separately sealed governance action); generic advisory, brainstorm, debate, QA, and worker requests can never reach Grok.

Workflow

Resolve the plugin root from this loaded file and read the exact request schema in <plugin-root>/README.md. Submit either:

• route=auto, action=architecture, governance=false, with exact Gemini, Codex, and Grok candidate rows; or
• route=grok, action=architecture, governance=false, with the exact Grok architecture row for an explicit target=grok request.

Use operation=execute, a bounded prompt, and the observed primary snapshot. When consulting on an authored design or plan, include its exact content and observed author model in the optional artifact snapshot so that family is also excluded. A frontmatter tier is not a coordinator field.

Scope

Eligible work includes:

• read-only codebase structure and dependency analysis;
• system and interface design;
• implementation sequencing and milestone plans;
• decomposition into bounded work units;
• risk, failure-mode, and tradeoff analysis;
• long-horizon coding strategy.

The architecture result is advice or a plan only. The trusted primary owns edits, shell/tests, integration, commits, PRs, merges, releases, deploys, and secrets. Never reinterpret "worker" language as execution authority.

Output contract

Ask for:

1. recommended architecture;
2. key invariants and threat boundaries;
3. implementation decomposition and dependency order;
4. verification plan;
5. unresolved decisions or assumptions.

If no eligible independent route is advertised, return typed unavailable. Do not reconstruct a raw Grok command or silently reuse a same-family reviewer.