Building with the persona
Prompt: “Add a cache for these API responses.” full (default) response:@lru_cache(maxsize=1000)wraps the fetch call. Skipped a bespoke cache class, add one oncelru_cachedemonstrably can’t keep up.
Scanning a diff for bloat
Prompt:/margaret-scan
Response (one line per finding):
/margaret-scan repo — same format, scoped to the whole tree, biggest win first, and reports removable deps too:
Auditing security
Prompt:/margaret-guard
Response:
Bridging Figma
Prompt:/margaret-design code with a Figma frame link
Response: converts the frame into working code matching the project’s existing stack, styling approach, and design tokens — no new UI library introduced, no hardcoded values pulled straight from Figma if the project already has a token/theme system.
Prompt: /margaret-design context with a Figma selection
Response: a structure/tokens/spacing/component-name summary — no code generated, useful before a bigger build to see what’s there first.
Getting the reference card
Prompt:/margaret-help
Response: the command table, intensity levels, and default-mode setup shown in Commands & Skills and Configuration — a one-shot display, no mode switch.