Apply the execution contract before this procedure.
Goal
Select the next procedure, load it, and continue the requested work within its gates.
Inputs / outputs
- In: the user request
- Out: the skill path to load, and nothing else
Reads / writes
- Read: this file,
directory.json,README.md - Write: none
Approval
None.
Route
Load only paths that exist in directory.json. No skill yet → Next skill: skills/rust-essentials (and skills/tdd if behaviour changes).
| If the request is… | Next skill |
|---|---|
| PRD / tickets / sprint | stop — agnostic-planning-skills |
| Existing crate, first action | skills/load-context |
| New or changed behaviour / TDD / bug | skills/tdd |
| Clone / lifetime / Arc | skills/ownership-borrowing |
| Newtype / enum state | skills/type-driven-design |
| unwrap / thiserror | skills/error-handling |
Anything .rs, including clap/tokio/unsafe/clippy until those skills exist |
skills/rust-essentials |
For existing crates, load load-context first. For behavior changes, route to tdd; for review or explanations, load only the relevant atomics. Load rust-essentials before any .rs write. Planning-only work routes to agnostic-planning-skills and does not open an implementation loop.
Steps
- Classify with the table
- First line:
Next skill: skills/<name> - Stop — do not write crate code.
Validation
No .rs in the crate was written in this turn.
Integration
Human catalog: README.md. Registry: directory.json. Browse: docs/reference/skill-catalog.md.