Reliable DSH plugin design starts with precise language and stable architectural boundaries. This section is the human-readable form of the installable dsh-plugin-dev skill: use the Context to name the platform correctly, then use the Decision Tree to choose one primary seam for each responsibility.
Install the skill
npx skills add BotHarness/dsh-skillBotHarness/dsh-skill mirrors the canonical skill folder in BotHarness/BotHarness. You can also copy the English Agent authority SKILL.md and its references/ into .agents/skills/dsh-plugin-dev/.
Stable foundation
| Page | Covers |
|---|---|
| Canonical context | DSH/Cordis leading words, native boundaries, and concept maps |
| Decision tree | Choosing Service, Event, Registry, SessionEvent, Projection, persistence, or execution seams |
| Release history | Independent Skill SemVer and the DSH revision verified by each release |
The skill intentionally excludes Host and Client API catalogs, Slot inventories, and community implementation patterns. Those details change quickly: after the Decision Tree selects a seam, verify its concrete mechanism against the current DSH official documentation, pinned source, and running Host.
Historical investigations remain in the canonical repository under docs/research/. They are evidence for maintainers, not published Skill guidance.
Product documentation is separate
This section names DSH/Cordis concepts only. BotHarness owns its product language and architecture under /dev: use the BotHarness domain glossary for product definitions and the runtime architecture for product relationships on DSH.
Provenance
Every page carries the Skill version, the DSH revision it was checked against, and the verification date. DSH is in developer preview; provenance makes the evidence inspectable, but it does not turn a concrete API detail into a permanent contract.