Agent skills,
scanned before you install them.
unictos is the phone-first store for AI agent skills, commands and tools. Every listing is security-scanned, pinned to the exact version we checked, and re-scanned whenever its code changes. Nothing here installs silently — ever.
Featured & verified
All skills
Verified Safe is a process, not a promise
Anyone can print a badge. Ours is a pipeline: scan, pin, re-scan — with the full report public on every listing. Here is exactly what happens before a skill earns the lime check.
Each skill is scanned for malicious patterns, hidden instructions and obfuscated code before it appears. High-risk findings block publication.
The install command points at the exact commit we checked — never a moving target. What we scanned is what you get.
Upstream repos are watched. A new commit triggers a fresh scan; if the score drops, the badge flips to “re-review pending” — never a silent update.
Score, findings, scan date and version history sit on every listing page. If we missed something, anyone can see it and report it in one tap.
The honest ceiling: automated scanning catches roughly two-thirds of malicious skills. A clean scan means “no known patterns found,” not “guaranteed safe” — that is why permissions are written in plain words and every install is a deliberate tap, never silent. unictos links to authors’ original repositories; we never wrap, modify or re-host their code.
Browse by category
Eight shelves, every listing on them scanned and pinned. Pick one to see its skills, or search the whole store from the bar above.
Highest rated right now
See the full catalog →Category
How Verified Safe works
A skill is code that runs inside your AI agent. Installing one from a bare link is like running a stranger’s script — so unictos scans everything first, pins the exact version, and keeps watching it. Here is the whole pipeline.
Every submitted skill passes a static security scan (SkillSpector) that hunts for malicious patterns: hidden instructions aimed at your agent, obfuscated or encoded payloads, credential-theft behaviour, destructive file operations and suspicious outbound calls. High-severity findings block the listing outright — the author fixes and resubmits.
Runs on every submission, $0 per scanSkills often point at scripts, prompts or data files hosted elsewhere. We fetch every referenced file, scan those too, and archive copies — so what the skill loads tomorrow is what we checked today, and a swapped file can’t sneak past.
Nothing loads uncheckedThe install command on every page resolves to one immutable commit hash — the precise code that passed the scan. Git tags and branch heads move; a pinned hash doesn’t. What we scanned is bit-for-bit what you install.
The scanned version is the installed versionNo vague “this skill may access your device.” Each listing spells out what the skill touches — files, network, scripts — in sentences a non-developer can judge in five seconds, straight from the scan report.
Readable before you tapWe watch each listed repo. A new commit re-triggers the full scan automatically; if the score falls or permissions change, the badge flips to “re-review pending” and the pinned install stays on the last clean version until a human clears it.
Never a silent updateAfter installing, people report whether the skill actually worked. Scans catch malice; the crowd catches junk. Both numbers sit on every listing — a high safety score with a low worked count tells its own story.
Two signals, not oneHonest limits
Automated scanning catches roughly two-thirds of malicious skills in published testing. “Verified Safe” means a skill passed this pipeline with no known-bad patterns — it is not a guarantee, and we will never phrase it as one. That is why permissions are public, installs are pinned and deliberate, and any listing can be reported in one tap. Security is a process we show our work on, not a badge we print.
Questions
List your skill on unictos
Free, forever. Your code stays in your own GitHub repo — unictos links to it, scans it, and pins installs to the version that passed. Fill this in and your skill enters the scan queue.
Queued for scanning
Your library
Skills you’ve saved from the catalog, with their pinned install commands. Saved on this device for this visit — sign-in and sync arrive with the full launch.