User custom Skills
User custom Skills are local, user-owned extensions. They do not enter this repository’s official skills/ directory merely because they were created with package guidance.
Lifecycle
create-custom-skill creates an isolated draft under ~/.weshop-skill-package/custom-skills/<slug>/ or another user-selected draft root.- The draft’s
intake.md records user ownership, intent, related Skill boundaries, execution behavior, safety, local check state, and proposed install target. It does not contain an external-source or license gate.
create-custom-skill inventories the draft, runs its bundled mechanical checker, performs semantic discovery/workflow/security self-checks, and repairs supported local-authoring issues.- A completed local check plus explicit user confirmation permits copying the runtime files into the active Agent’s user Skill directory. Exclude
intake.md from the installed runtime directory.
- A custom Skill remains unmanaged by the package updater. Refuse to overwrite an existing unmanaged directory.
The package checkout offers deterministic helpers:
weshop-skills custom init <slug>
weshop-skills custom check ~/.weshop-skill-package/custom-skills/<slug>
The initializer intentionally creates an incomplete draft. A check failure on its TODO placeholders is the expected state until authoring finishes.
Boundaries
- Use Importing an external Skill or project when the behavior comes from a third-party source. External intake records provenance and provider substitution before a native custom draft is reviewed.
- Use Adding or changing an Atom only when the user asks to upload or contribute the local Skill and maintainers decide to consider it for the official package. That transition starts independent intake and review; re-author and validate it as an official Atom rather than copying a user directory mechanically.
- Similarity never requires fusion. Every custom Skill names materially related installed Skills, records relationship scores, and states when each side wins.
- A static review does not authorize paid execution or prove a new WeShop route. New execution routes require current schema evidence and a separately authorized representative run before claiming readiness.