The body is lean: table-based scene routing, compact output templates, and no explanation of concepts Claude already knows; it assumes competence and avoids padding, with only minor repetition of the blockquote format.

It provides copy-paste-ready output templates (blockquote format and the six-field 目标/验收/动作/证据/状态/风险 execution layer) plus a concrete config mutation step, giving specific executable guidance rather than vague direction.

Sequencing is clear (immediate file loading, scene-routing table, set-default branch) and an explicit evidence/status checkpoint (candidate/needs_check/done_with_evidence) supplies the validation framing; no destructive/batch ops are present to cap it.

The body is an overview that signals one-level-deep references to real bundle files (methodology-ding.md, ding-reminders.md verified present) with clear navigation, content appropriately split rather than monolithic.

It names the domain ('Ding-style workplace reminders') but the only capability verb is 'reminders' — no list of multiple concrete actions, so it sits at 'names domain and some actions' rather than 'lists multiple specific concrete actions'.

It states what the skill does ('Ding-style workplace reminders rooted in...') and gives explicit 'when' guidance via the 'Triggers include:' clause and exclusion, satisfying both what and when without relying on implication.

An explicitly enumerated 'Triggers include:' list of ~30 niche terms plus a negative trigger ('Do not use for pure coding tasks') gives good coverage of the natural terms a user in this niche would actually say.

The highly specific Ding/钉 jargon (薛定谔的用户, 病态敏捷, 全景监狱) and the explicit exclusion of pure coding tasks carve a clear niche unlikely to trigger the wrong skill.