skill-incubator

Create and operate a durable, task-shaped HTML surface when a human needs to understand, inspect, edit, annotate, compare, decide, approve, or triage generated material with an agent. Use for visual review loops, decision sessions, media review, structured editors, reports, plans, batches, and other work that is clearer in an interactive surface than chat.

Run a documented grilling session in claudex with Claude leading the interview and the GPT/Codex-family grill-advisor challenging the opening design tree and auditing the final decisions. Use when the user wants Claude and Codex to stress-test a plan or design together while recording ADRs and domain terminology.

Compact the current conversation into a handoff document for another agent to pick up.

Watch a Quality Run Chapter execute in another session, read-only, and produce an evidence-first evaluation of how the agents behaved against the rules they were given. Use when a human wants a skill change, routing change, or orchestration change tested on a live Chapter and needs an independent record of what actually happened — not the Driver's own account of it.

Run /quality-loop under claudex with quota-aware Claude-spine and Codex-spine profiles. The selected family coordinates and implements while the other family supplies a persistent design advisor and fresh Correctness-vs-Design reviewer. Use for a substantial autonomous system build that needs cross-family scrutiny, explicit session boundaries, and resumable handoffs.

Turn an outcome, optional discovery brief, or existing GitHub Issue into an approved system design, autonomously implemented and adversarially reviewed semantic Chapters, and a human-readable System Story. Use when the user wants agents to build a substantial change while the human reviews architecture and behavior rather than every code diff, invokes /quality-loop, or resumes an active Quality Run.

Coordinate a lead agent, a persistent specialist or adapter, and a durable shared-state surface while keeping the lead thread thin. Use when work spans multiple feedback rounds, must survive compaction or session boundaries, or combines an agent-owned task with a delegated surface, reviewer, researcher, or advisor. Defines agent interaction and continuity; pair with model-orchestration for model selection and atelier for review surfaces.

How to combine the available models (Fable 5, Opus 4.8, Sonnet 5, Haiku 4.5, and GPT-5.6) by cognitive role on multi-step coding, research, design, and intensive audit tasks. Use when planning non-trivial work in a claudex or Codex proxy session, especially when deciding who gathers evidence, comprehends a large system, makes judgments, advises, implements, or independently reviews.

Check remaining subscription quota across providers (Claude, Codex) before routing or delegating token-heavy work, so a session sends bulk work to the model family with more headroom. Use at session start and before large delegations or fan-outs.

Make human-facing text clear, natural, and appropriate for its readers without losing technical depth or important meaning. Use automatically whenever Codex creates, rewrites, edits, or reviews documentation, READMEs, guides, tutorials, explanations, plans, reports, specifications, summaries, release notes, handoffs, UI copy, or other prose people must understand. Use alongside file-format or documentation skills. Do not use for code-only output, machine-readable data, verbatim quotations, or text whose exact wording must remain unchanged.