メニュー
Diagnoses hard bugs and performance regressions with a controlled feedback loop, and implements the fix only when requested. Use when the failure is flaky, not yet reproducible, still lacks a root cause, or needs measurement.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Structured design critique across usability, hierarchy, consistency, and accessibility. Use when the user shares a Figma link, screenshot, mockup, or asks for design feedback ("review this design", "critique this screen", "what do you think of this UI?"). Use after design-router when a theme has been selected and applied.
Runs a disciplined ablation workflow for MixJam Electron (MJE) bugs to prove which changed files or layers are actually required for a fix. Use when an MJE import, playback, or UI bug has a known pass/fail check but several edits may be involved, or when the user asks which change actually fixed it, what can be removed, or for the smallest proven fix.
Creates or updates repository specs, acceptance criteria, durable decisions, and documentation for new work. Use when adding a feature or defining an ambiguous slice, when making architectural decisions or changing contracts (API, persistence, import, playback), or when recording durable context (glossary, conventions, trade-offs) for future engineers and agents.
Audits the MixJam Electron (MJE) codebase for dead TypeScript code, orphan files, and unused symbols across main and renderer processes, triages findings into live dead code or false positives, and optionally removes provably dead code with focused validation. Use for a dead-code audit, or for cleanup explicitly requested from analyzer findings.
Routes a requested UI visual direction to one bundled design theme. Use when building or restyling an interface that needs an explicit aesthetic.
Remove AI slop from every file in the repository while preserving behavior and genuine information.
| name | diagnose |
| description | Diagnoses hard bugs and performance regressions with a controlled feedback loop, and implements the fix only when requested. Use when the failure is flaky, not yet reproducible, still lacks a root cause, or needs measurement. |
A discipline for hard bugs. When the failure mode and repro are already narrow, use the smallest feedback loop that can still falsify the suspected cause.
Work through these in order:
improve-codebase-architecture when warranted.The successful diagnosis branch is complete when:
The blocked branch is complete when the feedback loop cannot be built after the documented attempts, the missing artifact or access is named, and no root cause or fix is claimed.
Use REFERENCE.md for full phase instructions, loop construction patterns, and the HITL script at scripts/hitl-loop.template.ps1. Use EXAMPLES.md for concrete diagnosis loops.