styleseedSAFE

Open-source design-method engine for Claude Code, Codex & Cursor. 23 agent skills for fixed design judgment, multiple grammars, semantic palettes, reference compilation, and evidence-verified UI. MIT.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: styleseed
description: Build, improve, or inspect UI with StyleSeed. Preserves project design choices and routes setup, layout, tokens, review, rendered verification, and updates through one entry point.
---

# StyleSeed

Use the user's goal to choose one first workflow below. This is the only registered core
skill; the 22 workflows and their executable dependencies ship inside this directory.
Read only the selected workflow, then references needed for that task.

## Registry-first artifact boundary

When `.styleseed/project.json` and `.styleseed/artifacts/index.json` exist, resolve the requested
artifact first. Read its `.styleseed/bundles/<artifact-id>.md` and manifest. A partial or invalid
registry is an error; never fall back to the global legacy bundle. Legacy projects may use
`.styleseed/effective-rules.md` only when no registry exists.

Before the first workflow in a project/task session, follow the
[update preflight](workflows/ss-update/references/update-preflight.md).

## Working contract

- Preserve approved project design choices. Support expert judgment; a score is not human acceptance.
- Resolve the current artifact first. For multiple plausible artifacts, ask one bounded clarification question.
- Never fan out to run every workflow. Choose exactly one first workflow; it may call the required gates.
- Honor the user's authorized scope. Audit, score, verify, and status requests do not authorize edits.
  Workflow metadata records the intended tool/scope boundary even though it is not host registration metadata.
- Paths in workflow files are relative to those files. Resolve installed scripts from this skill's
  actual directory, never assume a particular agent folder or download a missing sibling skill.
- In older instructions, `/ss-build`, `$ss-build`, and other `ss-*` names mean the corresponding
  internal workflow. Read `workflows/<name>/WORKFLOW.md`; do not invoke an unregistered skill.
  Users may say `StyleSeed build`, `$styleseed build`, or the old workflow name in natural language.
- Learning is an optional extension, not part of the core install. Only for explicit learning capture,
  use separately installed `ss-learn`; if absent, report that dependency and never auto-install it.

## Choose exactly one first workflow

| Request | Read |
|---|---|
| Build or redesign a screen, page, or component | [build](workflows/ss-build/WORKFLOW.md) |
| First setup or missing design lock | [setup](workflows/ss-setup/WORKFLOW.md) |
| Explore creative directions; direction not yet selected | [studio](workflows/ss-studio/WORKFLOW.md) |
| Use screenshot, URL, Figma export, or existing UI as a reference | [reference](workflows/ss-reference/WORKFLOW.md) |
| What is wrong with this screen; UX critique | [audit](workflows/ss-audit/WORKFLOW.md) |
| Review implementation against project design rules | [review](workflows/ss-review/WORKFLOW.md) |
| Score implementation or run code/evidence gates | [score](workflows/ss-score/WORKFLOW.md) |
| Inspect rendered output, screenshot, or spacing in pixels | [verify](workflows/ss-verify/WORKFLOW.md) |
| Change spacing/density or another single design axis | [dial](workflows/ss-dial/WORKFLOW.md) |
| Apply an aesthetic profile | [restyle](workflows/ss-restyle/WORKFLOW.md) |
| Brand color, semantic palette, or design tokens | [tokens](workflows/ss-tokens/WORKFLOW.md) |
| Named animation or motion treatment | [motion](workflows/ss-motion/WORKFLOW.md) |
| Specifically scaffold a page | [page](workflows/ss-page/WORKFLOW.md) |
| Specifically generate a reusable component | [component](workflows/ss-component/WORKFLOW.md) |
| Compose existing primitives into a UI pattern | [pattern](workflows/ss-pattern/WORKFLOW.md) |
| Accessibility audit and requested fixes | [a11y](workflows/ss-a11y/WORKFLOW.md) |
| Quick design lint | [lint](workflows/ss-lint/WORKFLOW.md) |
| Navigation and user flows | [flow](workflows/ss-flow/WORKFLOW.md) |
| UX microcopy | [copy](workflows/ss-copy/WORKFLOW.md) |
| Loading, empty, success, or error states | [feedback](workflows/ss-feedback/WORKFLOW.md) |
| Update installation or consolidate old skill entries | [update](workflows/ss-update/WORKFLOW.md) |
| Compile rules, inspect local install/bundle/evidence health | [resolve](workflows/ss-resolve/WORKFLOW.md) |

For status-only requests use resolve's read-only `scripts/styleseed-doctor.mjs`; do not update,
recompile, or visually inspect instead. For vague requests infer the first step from project state
and the user's goal; ask only when an unresolved choice materially changes the work.

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

declared (5 observation(s))

Network

declared (1 observation(s))

Shell

declared (3 observation(s))

Dependencies

pinned

Secrets in source

none-found

Findings (6)

.agents/skills

resolve(scriptDir, "../../../../color/generator.mjs"),

resolve(scriptDir, "../../../engine/color/generator.mjs"),

import { sourceInventory } from '../../ss-score/scripts/evidence-contract.mjs';

const { verifyEvidenceRun } = await import("../../ss-score/scripts/evidence-gate.mjs");

} from "../../ss-resolve/scripts/runtime-contract.mjs";

Gates applied: no_behavioural_pass.