Frontend-craft — the design chain around UI work

Generic AI-looking UI is the failure mode; this chain is the antidote. It wraps the implementation with project-aware design context first, and a critique/polish pass after. The chain matters more than any single link — run it in order.

No UI surface in this task? Exit silently — do not run the chain.

Before building

1. Load the design ground truth yourself (don't assume the sub-skills will): the repo's design tokens (e.g. saferskills ui/styles/tokens.css), a .impeccable.md if the repo has one, the repo's .claude/rules/design-system.md, and the vault's 03-Product/Brand and Design System/DESIGN.md.
2. Skill: `impeccable:impeccable` — the design anchor (gathers its own project context; your step 0 reading lets you correct it where it misses repo specifics).
3. Skill: `frontend-design:frontend-design` — distinctive interface generation that avoids generic AI aesthetics.
4. Skill: `frontend-design-animation` — purposeful motion: transitions and micro-interactions, elegant not gratuitous.

Then build, with a strict bias to reuse: design-system components and tokens before hand-rolled anything. Hardcoded colors/sizes where a token exists is a defect.

After building

1. Skill: `impeccable:critique` — evaluate what was built.
2. Skill: `impeccable:polish` — final quality pass on the findings.

Handoff discipline

Invoking a sub-skill buries the parent workflow's next step under fresh output — that's the known failure mode of chains. So after EVERY link, restate position explicitly: "→ frontend-craft: finished frontend-design (2/5), next is animation (3/5)." When the chain completes, hand back to the calling workflow with the same restate so the parent sequence resumes where it left off.