iOS still makes you tap through an “Allow Untrusted Shortcuts” warning before running anything you didn’t build yourself, but that warning never explains what the shortcut actually does. Trusty Trail closes that gap: it reads a shortcut’s actions before you install it and has AI translate them into a plain-language safety report.
What it does
Once you point Trusty Trail at a shortcut, it converts it into a property list, the format the Shortcuts app itself reads, and scans every action for suspicious behavior: deleting data, sending data to an outside server, silently changing device settings. Matches get sorted into twenty categories, each flagged red, orange, or yellow by risk.
Two AI models then write up what was found. Google’s Gemma drafts the explanation, and a pass through Cerebras’s GLM cleans up the wording before you see the finished report in a light or dark VCard-style menu.
What you’ll need
- An internet connection: the AI step sends shortcut data out to Google’s and Cerebras’s APIs, so a check won’t finish offline.
- A shortcut to check: pick one on your device, search inside Trusty Trail, or hand it a link from iCloud, RoutineHub, or a Reddit post.
- “Allow Sharing Large Amounts of Data” turned on: under Settings > Apps > Shortcuts > Advanced, if asked.
Adding it to your iPhone
- Run it once and approve the permissions it asks for, including network access.
- If a prompt for “Allow Sharing Large Amounts of Data” appears, turn it on under Settings > Apps > Shortcuts > Advanced.
First run
The main menu gives you three ways into a check:
- Pick from your own list of installed shortcuts.
- Search by a shortcut’s name.
- Paste in a URL, whether it’s from iCloud, RoutineHub, or RoutineHub’s direct download page.
Importing also works from the Share Sheet. Share a Reddit post or comment with shortcut links inside, and Trusty Trail lists every one it finds, with name, icon, and poster attached.
Using it day-to-day
After a check finishes, Trusty Trail shows a verdict up top, safe or not, then a breakdown of every category where something matched. Tap into any category to read the AI’s explanation of what the action does and why it did or didn’t raise a flag. A clean result unlocks a “Generate a safety badge” button, saved as Markdown, an image link, or straight to Photos, and every badge carries details of which shortcut was verified so nobody can pass one off as their own.
Tips
- Treat the AI verdict as a strong signal, not gospel. Language models occasionally get details wrong, so a “safe” result deserves a second look on anything touching your data.
- Nuvole AI can freeze on launch. If the text field never shows up, close it and try again.
- Very large shortcuts hit a size ceiling. Get close and the report warns the description might be off; go past it and that step gets skipped entirely.
Quick answers
Does Trusty Trail actually run the shortcuts I’m checking?
Not in the sense of executing them. It converts a shortcut into a Property List, the structured format the Shortcuts app reads internally, and checks which actions are present. That’s static analysis, not execution.
What happens if a shortcut is too big for the AI to check?
The suspicious-action list still comes through, since that part never depends on AI. What gets skipped, or flagged as possibly inaccurate, is the written explanation, so a huge shortcut still gets checked, just with a shorter report.
Can I trust a shortcut just because it has a safety badge?
Treat it as a reasonable signal, not a guarantee. A badge only gets generated after a live check that comes back safe, and it embeds details of which shortcut was verified so it can’t be reused elsewhere. The check underneath still runs on AI models, though, and those can misjudge intent like any automated scanner.
First time adding a shortcut from a website? Here's what each prompt means.
Version history
| Version | Date | Changes |
|---|---|---|
| 3.2 | Sep 2026 | — Badge Generation Improvements: • Introduced a Universal badge style crafted to look seamless in both Light and Dark modes. • Addressed color-detection inaccuracies when generating highlight badges from shortcut icons. — Numerous Fixes: • Restored shortcut verification from RoutineHub links. • Improved shortcut icon rendering when inspected from URL. • Corrected false-positive "unsafe" AI verdicts on safe shortcuts. • Fixed a bug causing rogue menus or unexpected crashes when tapping warning and error alerts during AI analysis. • Prompts have been significantly improved. • Enhanced backend stability. |
| 3.1.1 | Jul 2026 | — Significant AI Checking Update: • Now, during AI checks, Trusty Trail uses the faster Google Gemma 4 31B instead of the slow GLM-5. • The prompts have been updated. • Backend stability significantly improved. |
