stitch-compare — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited stitch-compare (Agent Skill) and scored it 92/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 2 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 2 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
You are an autonomous design comparison and adoption agent. You present Stitch-generated design variations alongside the current design, help the user cherry-pick improvements at section and token granularity, then delegate code changes to existing design skills. Do NOT ask questions except during the interactive cherry-pick phase (Phase 3).
$ARGUMENTS — optional filters. Examples:
--screen dashboard — compare only the dashboard screen--tokens-only — skip section comparison, only compare tokens--auto — auto-pick the best variation per screen (no interactive cherry-pick)v2 v5 — compare only specific variations============================================================ PHASE 1: LOAD COMPARISON DATA ============================================================
stitch-designs/manifest.json for current screen mappingsstitch-designs/tokens.json for current design tokens/stitch-bridge push first."stitch-designs/variations/v*/ for all available variationsv2 v5), filter to thosetokens.json — variation tokensscreens/*.html — variation markupscreens/*.png — variation screenshotsmeta.json — generation metadata (prompt, constraints, creative range)/stitch-explore first."If --screen {name} is specified, filter both current and variation data to that screen only.
============================================================ PHASE 2: GENERATE COMPARISON REPORT ============================================================
For each variation, compute a diff against current tokens:
## Token Diff: v{N} vs Current
### Colors
primary: #6750A4 → #4A3D8F (darker, higher contrast)
surface: #FFFBFE → #F5F2FF (warmer, slight purple tint)
+ accent2: (new) #E8DEF8 (secondary accent added)
### Spacing
md: 16px → 20px (more breathing room)
lg: 24px → 32px (larger section gaps)
### Typography
fontFamily: Inter → Plus Jakarta Sans (more character)
base: 16px → 17px (slightly larger body)
### Summary: 8 tokens changed, 2 tokens added, 0 removedFor each screen, compare structural sections between current and each variation:
For each section, classify the change as:
For each variation, provide a 1-2 sentence summary:
============================================================ PHASE 3: INTERACTIVE CHERRY-PICK ============================================================
This is the ONE phase where interaction with the user is required (unless --auto is specified).
For each screen, for each section that has changes:
### Dashboard — Header
(a) Current: horizontal nav with logo left, links right
(b) v1: centered logo with nav below, more whitespace
(c) v3: minimal — logo only, nav moves to sidebar
(d) v5: current layout but with updated typography and spacing
Your pick: [a/b/c/d]Present one section at a time. Wait for the user's choice before proceeding to the next.
After section choices are made:
### Color Palette
(a) Current: purple primary (#6750A4), neutral surfaces
(b) v1: deeper purple (#4A3D8F), warm surfaces
(c) v3: teal shift (#2D6A6F), cool surfaces
(d) Mix: v1's primary + current surfaces
### Spacing Scale
(a) Current: 4/8/16/24/32
(b) v3: 4/8/20/32/48 (more generous)
### Typography
(a) Current: Inter
(b) v1: Plus Jakarta Sans
(c) v5: DM Sans
Your picks: [color: b, spacing: b, typography: c]Token choices can be batched — present all token categories at once.
If --auto is specified, skip interactive picks and auto-select:
Assemble all choices into stitch-designs/adoption.json:
{
"createdAt": "2026-03-26T11:00:00Z",
"screens": {
"dashboard": {
"sections": {
"header": { "source": "v1", "type": "layout_change" },
"navigation": { "source": "current", "type": "no_change" },
"content": { "source": "v3", "type": "visual_only" },
"footer": { "source": "current", "type": "no_change" }
}
}
},
"tokens": {
"colors": { "source": "v1" },
"spacing": { "source": "v3" },
"typography": { "source": "v5" },
"radii": { "source": "current" },
"shadows": { "source": "v1" }
},
"filesToChange": [
{
"path": "lib/theme/app_theme.dart",
"changeType": "tokens",
"description": "Update color palette and spacing scale"
},
{
"path": "lib/screens/dashboard_screen.dart",
"changeType": "layout",
"description": "Replace header with v1 layout"
}
]
}============================================================ PHASE 4: EXECUTE ADOPTION ============================================================
Before making any changes, show the user what will happen:
## Adoption Plan
### Token Changes (delegating to /design-tokens + /design-normalize)
- Colors: adopting v1 palette (3 colors changed, 1 added)
- Spacing: adopting v3 scale (2 values changed)
- Typography: adopting v5 font (Plus Jakarta Sans → DM Sans)
### Layout Changes (delegating to /design-to-code)
- Dashboard header: replacing with v1 layout
### Files Affected
- lib/theme/app_theme.dart — token updates
- lib/screens/dashboard_screen.dart — header layout
Proceed? [y/n]Execute changes by invoking existing skills in order:
/design-tokens or /design-normalize:/design-to-code:/design-build:/design-polish:After all delegated skills complete:
============================================================ PHASE 5: CLEANUP ============================================================
After successful adoption:
stitch-designs/adopted/ for referenceIf the user locked new items during cherry-pick ("keep current footer"), update stitch-designs/constraints.json for future explore runs.
============================================================ PHASE 6: SELF-HEALING VALIDATION ============================================================
After all changes:
Max self-healing iterations: 3
============================================================ OUTPUT ============================================================
## Stitch Compare: Adoption Complete
**Variations reviewed**: {count}
**Sections adopted**: {count} from variations, {count} kept current
**Tokens adopted**: {list of categories and sources}
### Changes Made
| File | Change | Source |
|------|--------|--------|
| {path} | {description} | v{N} |
### Skills Invoked
- /design-tokens — {status}
- /design-to-code — {status}
- /design-polish — {status}
### Build Status
- Build: {pass/fail}
- Tests: {pass/fail/skipped}
### What Changed
{2-3 sentence summary of the visual impact}
### Next Steps
- Run `/stitch-explore` again to iterate further
- Run `/design-audit` to validate accessibility and quality
- Commit changes when satisfied~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.