Skill Title
One sentence defining the repeatable procedure and the decision it owns.
Use When
- Bullet the concrete situations that should trigger this skill.
Do Not Use When
- Bullet the situations where another skill is the better match (name it).
Required Inputs
| Artefact or context | Source | Required? | Why |
|---|---|---|---|
| Named input | User or upstream skill | yes | Decision it enables |
Workflow
- Step-by-step, citing
doctrine/references/*where a canonical rule applies.
Decision Rules
| Condition | Action | Wrong-choice failure |
|---|---|---|
| Concrete threshold or state | Required response | Consequence to avoid |
Capability Contract
- State the minimum read, search, edit, execute, network, render, or delegation capabilities.
- Default reviews and audits to read-only. Require explicit authority before mutation or publication.
Degraded Mode
- State the narrowest useful output when an input, renderer, browser, font, licence, or test is unavailable.
- Name unverified checks and block release when the missing evidence is material.
Anti-Patterns
- Name at least five concrete mistakes and pair each with its correction.
Outputs
| Artefact | Consumer | Evidence and acceptance condition |
|---|---|---|
| Named output | Downstream role or skill | Observable proof of completion |
Examples
- See
examples/— every craft skill MUST ship ≥1 worked example (applied spec / before-after / sample artifact / worked decision). Never lorem. SeeCONTRIBUTING.md.
References
doctrine/design-doctrine.mdand the specificdoctrine/references/*this skill relies on (incl.wcag-2.2-criteria.mdandweb-performance-budgets-2026.mdfor UI/web skills).