svvy

The concierge proxy — turn a vague user script ("I need the agent to help me do X") into a context contract, route it to the right skills/workflows, add backpressure (tests/evals/reviews/approvals), execute, and report. Use when a request is multi-step, durable, or human-in-the-loop and you'd otherwise hand-roll the orchestration; skip it for a single prompt → single answer.

Turn acceptance criteria into a runnable Smithers eval suite (JSONL cases + rubric) and wire it to `smithers eval`. Use when a workflow's quality must be measured and regression-tested — not "looks good" once, but a repeatable check that fails when the model OR the harness regresses.

Turn a run's persisted state into a self-contained HTML slideshow report — objective, decisions, graph, gates, evals, artifacts, failures, and the recommended next run. Use when a long or important run needs a legible, shareable summary instead of scrolling raw logs.

Design the Zod output schema of a Smithers <Task> as the contract between steps. Use when a step's output feeds a later step (or a branch/loop condition) and must be reliable — design the schema first, keep it minimal, and prefer typed fields over prose so downstream rendering can depend on it.

Drive Smithers, a durable control plane for long-running coding agents. Use when the user wants multi-step, long-running, crash-safe, or human-in-the-loop agent work: "orchestrate agents", "run a workflow", "implement this and review it", "keep iterating until tests pass", "plan then build", or anything that needs retries, approvals, replay, or evals across multiple AI steps. YOU (the agent) run Smithers on the user's behalf; it is not a GUI the human clicks. You are an ORCHESTRATOR: run long-running, multi-step, or background work *through* Smithers, not through your own ad-hoc subagents; spend your time observing the run and reporting.

Create a single-file, executable proof of concept for a proposed feature, architecture change, state model, workflow, or runtime lifecycle. Use when the user wants a runnable POC that is easy to read with zero prior context, prefers inline types and top-down structure, minimizes abstraction and scrolling, and is intended to be reviewed and refined collaboratively after the first draft.

Manually sync a fully resolved note from docs/draft-notes.md into docs/prd.md and docs/progress.md, then remove the absorbed draft note and commit the docs change. Use only when explicitly invoking this workflow.

Smooth Brain Explain: re-explain your last output (or respond to an accompanying prompt) in incredibly simple, concise terms readable by a lazy human who understands the system but doesn't like reading or unnecessary jargon.