メニュー
Use for /he:learn; durable guard after repeated misses, review loops, or workflow gaps.
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
| name | he-learn |
| description | Use for /he:learn; durable guard after repeated misses, review loops, or workflow gaps. |
Stage 5/5. Use /he:learn only when he-state.json contains open learning findings after repeated misses, review gaps, or ship findings. Finish the durable guard; do not timebox it.
Read ../workflow-help/references/route-map.md, ../repeated-failure-learning/SKILL.md, and ../skill-creator/SKILL.md before acting.
he-state.json before/after each learning step; validate it before loop-complete yes/he:ship should close with Next: loop complete: yesskill-creator only for skill/stage-contract changeshe-learn until the durable guard exists and passesNext: loop complete: yes/no. No transcript dump; future runs can resume from stateUse for /he:plan; stage 1 readiness with Treehouse, Grill Me, owner, proof path, PASS/FAIL.
Use for UI components, design systems, atomic design, tokens, themes, hardcoded visuals, or missing design SSOT.
Use for module boundaries, public APIs, ownership, abstractions, wrappers, test seams, or architecture review.
Use for visual PDFs: create, polish, export, verify decks, reports, proposals, one-pagers, or HTML-to-PDF.
Use for hard bugs, failing behavior, flaky tests, exceptions, regressions, or unexplained slowness before fixing.
Use for real UI E2E smoke/regression proof across web, Flutter, browser, Chrome, or device flows, with artifacts.