f00215a5 UI diagrams
This skill is an orchestration boundary for a parent UI-manual workflow. It never creates, previews, exports, or changes diagram artifacts.
Route the request
- Confirm that the requested visual is a diagram rather than a screenshot. For a screenshot request, return control to the caller without invoking draw.io.
- Run
scripts/check_readiness.pyand read its JSONstatus:ready,needs-install, orunavailable.
When dependencies need installation
For needs-install, read dependency-and-install-policy.md. Present the missing components, exact platform command, upstream source URL, install scope, and this exact question: 是否同意為本次任務安裝？ Do not run an installer unless the user gives 明確同意 in this current task and the runtime grants the required approval.
On an approved install, clone https://github.com/Agents365-ai/drawio-skill.git into the current user's .codex/skills/drawio-skill only when it is missing. Run only the policy-supported platform installer, then re-run readiness. If it is ready, use the policy's current-task direct-load procedure and immediately hand off; never generate output while loading the downstream skill.
If clone, desktop installation, the recheck, downstream skill loading, or handoff fails, follow the policy's terminal failure path. Do not restart the consent or installation flow.
If the user declines, or readiness is unavailable, report that status only in chat, explicitly return the caller to the continuing UI manual workflow, and 只停止圖表分支. Do not attempt another installer, diagram tool, artifact operation, or workaround.
Handoff boundary
When readiness is ready, 立即交棒給 $drawio-skill. Do not wait to generate anything and 不自行建立、預覽、匯出或修改 .drawio or PNG files. Downstream owns every generation and delivery decision; its default delivery is diagram.drawio plus diagram.drawio.png, which the user may explicitly override before downstream generation.