stitch-pipeline — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited stitch-pipeline (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 improvement pipeline. You chain stitch-bridge, stitch-explore, and stitch-compare to take an existing app's design, generate alternatives in Google Stitch, and help the user adopt improvements. Do NOT ask questions except during the cherry-pick phase of stitch-compare.
$ARGUMENTS — improvement prompt and optional flags. Examples:
"make the onboarding feel less overwhelming" — full pipeline with preset prompt--screen settings "modernize the layout" — scope to one screen--skip-push — skip bridge push (already synced)--skip-explore — skip explore (variations already generated)--variations 5 — generate 5 variations instead of default 3--creative-range reimagine — radical exploration--auto — auto-pick best variations (no interactive cherry-pick)============================================================ PHASE 1: PRE-FLIGHT ============================================================
stitch MCP server is connectedstitch-designs/manifest.json exist? → can skip pushstitch-designs/variations/v*/ exist? → can skip explorestitch-designs/adoption.json exist? → previous adoption in progress, warn userBased on flags and existing state:
| Flag | Push | Explore | Compare |
|---|---|---|---|
| (none) | yes | yes | yes |
--skip-push | skip | yes | yes |
--skip-explore | skip | skip | yes |
| manifest exists + no flag | skip | yes | yes |
============================================================ PHASE 2: BRIDGE PUSH ============================================================
Skip if --skip-push or manifest already exists and is recent (< 1 hour old).
Follow the instructions defined in the /stitch-bridge skill with arguments: push
Pause point: After push completes, report what was synced before continuing.
Synced {count} screens to Stitch project {id}.
Tokens extracted: {count} colors, {count} spacing, {count} typography.
Continuing to explore...============================================================ PHASE 3: EXPLORE ============================================================
Skip if --skip-explore or variations already exist.
Follow the instructions defined in the /stitch-explore skill with arguments from $ARGUMENTS:
--variations count if specified--creative-range if specified--screen filter if specifiedPause point: After explore completes, report variations before continuing.
Generated {count} variations.
{1-line summary of each variation}
Continuing to compare...Iteration loop: If explore results look weak (all variations are near-identical, or none meaningfully address the prompt):
--creative-range reimagine============================================================ PHASE 4: COMPARE AND ADOPT ============================================================
Follow the instructions defined in the /stitch-compare skill with arguments from $ARGUMENTS:
--auto if specified--screen filter if specifiedThis phase is interactive (presents choices to the user) unless --auto is specified.
============================================================ PHASE 5: POST-ADOPTION ============================================================
stitch-designs/variations/stitch-designs/adopted/ for referenceconstraints.json for future runsmanifest.json updatedBased on what was adopted, suggest relevant existing skills:
/design-adapt for responsive verification/design-audit for accessibility check/design-polish for consistency pass/design-animate============================================================ PHASE 6: SELF-HEALING VALIDATION ============================================================
If any phase fails:
If the pipeline is interrupted and re-run:
stitch-designs/Max self-healing iterations: 3 per phase
============================================================ OUTPUT ============================================================
## Stitch Pipeline: Complete
**Prompt**: "{improvement prompt or 'auto-detected'}"
**Creative range**: {REFINE|EXPLORE|REIMAGINE}
### Pipeline Summary
| Phase | Status | Details |
|-------|--------|---------|
| Bridge Push | {done/skipped} | {count} screens synced |
| Explore | {done/skipped} | {count} variations generated |
| Compare | {done} | {count} sections adopted, {count} kept |
| Build | {pass/fail} | {details} |
### What Changed
{3-5 sentence summary of the visual transformation — what the app looked like before and what it looks like now}
### Adopted From
- Tokens: {sources per category}
- Sections: {sources per screen}
### Files Modified
| File | Change |
|------|--------|
| {path} | {description} |
### Suggested Follow-ups
- {skill suggestion based on what changed}
- {skill suggestion}
### Re-run Options
- `/stitch-pipeline --skip-push "try different direction"` — keep current sync, explore new ideas
- `/stitch-explore "specific prompt"` — generate more variations without full pipeline
- `/stitch-compare` — re-compare if you want to change your picks~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.