architect

Generate an interactive HTML architecture diagram (`diagram.html`) with drill-down navigation and Comment Mode. If architecture artifacts already exist, render from them; if not, first run architect-init to discover the repo and create the artifacts.

Architecture planning sidecar for Claude's native Plan Mode. Turn product requirements, constraints, team size, timeline, assumptions, and tradeoffs into canonical architecture artifacts and, when useful, an interactive diagram. Strongly consider this skill in Plan Mode when a user is planning a new or changing software system and asks for the architecture, system design, service breakdown, or technical plan before implementation, including prompts like "propose the architecture", "design the system", or "help me plan the architecture."

Generate an interactive HTML architecture diagram (`diagram.html`) with drill-down navigation and Comment Mode. If architecture artifacts already exist, render from them; if not, first run architect-init to discover the repo and create the artifacts.

Architecture planning sidecar for Claude's native Plan Mode. Turn product requirements, constraints, team size, timeline, assumptions, and tradeoffs into canonical architecture artifacts and, when useful, an interactive diagram. Strongly consider this skill in Plan Mode when a user is planning a new or changing software system and asks for the architecture, system design, service breakdown, or technical plan before implementation, including prompts like "propose the architecture", "design the system", or "help me plan the architecture."

Discover and model the architecture of an unfamiliar codebase, emit a canonical architecture model plus derived structured view files for multiple audiences, then automatically hand off to architect-diagram for the interactive HTML diagram. Use this skill when asked to explore a repository, identify runtime boundaries, data ownership, key workflows, or generate architecture artifacts from source code.

Discover and model the architecture of an unfamiliar codebase, emit a canonical architecture model plus derived structured view files for multiple audiences, then automatically hand off to architect-diagram for the interactive HTML diagram. Use this skill when asked to explore a repository, identify runtime boundaries, data ownership, key workflows, or generate architecture artifacts from source code.

Generate the secondary Claude Imagine upload bundle (`diagram-prompt.md`) from existing architecture artifacts and rendered diagram output. Use when explicitly requested.

Run a repeatable eval round for the architect-init skill. Use this when you want to evaluate a new version of the skill on a fresh codebase, get subagent review of the generated architecture artifacts, reflect on the run, and optionally apply improvements to architect-init after user approval.

Generate a planning-time architecture proposal (before implementation) from user intent, constraints, assumptions, and tradeoffs; iterate with engineer feedback until explicit approval; emit canonical architecture artifacts using the shared schema. Use when the user is designing a new system or major feature and wants architecture steering without repo discovery.

Generate the secondary Claude Imagine upload bundle (`diagram-prompt.md`) from existing architecture artifacts and rendered diagram output. Use when explicitly requested.

Run repeatable planning-mode evals for architect-plan using an iteration workspace pattern with with_skill vs baseline runs, three-lane grading (contract, semantic, steering-loop), and explicit promotion gates.