Findings & checks · 2 flagged
HIGHLong base64-encoded blob hidden in the skill documentationSS-SKILL-INJECT-B64-PAYLOAD-01 · Prompt injection · README.md
A base64 string of 128+ characters appears in a documentation file. Encoded prompt injection hides the hostile instruction in base64 — invisible to keyword filters — and relies on the agent's ability to decode it at runtime. There is no normal authoring reason to embed a multi-hundred-byte base64 blob in skill docs.
- Decode the base64 string and confirm it is not an instruction directed at the agent.
- Move any legitimate binary or signature data into a dedicated file (
*.sig,SIGNATURES) outside the documentation.
Trace & refs
MEDIUMInstruction telling the agent not to ask for approvalSS-SKILL-INJECT-DONT-ASK-01 · Prompt injection · evals/ios-accessibility/ios-accessibility-workspace/iteration-2/wcag-auto-vs-manual/with_skill/outputs/response.md
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.
- Delete blanket "don't ask / no need to confirm" directives from the skill.
- If the skill is a genuine autonomous job, restrict the opt-out to a named non-destructive action rather than all actions.
Trace & refs
Every scanned point with the score it earned and what moved between them.
Aggregate score unchanged between these scans.
The primary manifest — the file an agent reads to learn what this artifact does.
<p align="center"> <img src="assets/banner.png" alt="iOS Agent Skills" width="100%"> </p>
<p align="center"> <a href="https://github.com/rusel95/ios-agent-skills"><img src="https://visitor-badge.laobi.icu/badge?page_id=rusel95.ios-agent-skills&left_color=gray&right_color=blue&left_text=visitors" alt="visitors"/></a> <a href="https://github.com/rusel95/ios-agent-skills"><img src="https://img.shields.io/github/stars/rusel95/ios-agent-skills?style=flat-square&color=yellow" alt="stars"/></a> <img src="https://img.shields.io/badge/skills-11-purple?style=flat-square" alt="skills"/> <img src="https://img.shields.io/badge/models_tested-3-teal?style=flat-square" alt="models"/> <img src="https://img.shields.io/badge/assertions-850%2B-orange?style=flat-square" alt="assertions"/> <img src="https://img.shields.io/badge/scenarios-260%2B-green?style=flat-square" alt="scenarios"/> <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-brightgreen?style=flat-square" alt="license"/></a> <a href="https://lobehub.com/skills?q=rusel95-ios-agent-skills"><img src="https://img.shields.io/badge/LobeHub-Skills_Marketplace-blue?style=flat-square&logo=data:image/svg+xml;base64,PHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHZpZXdCb3g9IjAgMCAyNCAyNCI+PHBhdGggZD0iTTEyIDJDNi40OCAyIDIgNi40OCAyIDEyczQuNDggMTAgMTAgMTAgMTAtNC40OCAxMC0xMFMxNy41MiAyIDEyIDJ6IiBmaWxsPSIjZmZmIi8+PC9zdmc+" alt="LobeHub"/></a> </p>
iOS Agent Skills
The first and most comprehensively benchmarked iOS skill marketplace for Claude Code, Codex, and 40+ AI coding tools.
11 enterprise-grade skills covering architecture, concurrency, testing, security, accessibility, and localization — every skill benchmarked with discriminating assertions and blind A/B quality scoring across multiple LLMs. No other iOS skill collection has this level of rigorous, reproducible evaluation — 850+ assertions across 260+ scenarios, tested on Claude Sonnet 4.6, GPT-5.4, and Gemini 3.1 Pro.
Benchmark Results
Every skill is benchmarked against multiple LLMs with discriminating assertions and blind A/B quality scoring.
| Skill | Sonnet 4.6 | GPT-5.4 | Gemini 3.1 Pro |
|---|---|---|---|
| swiftui-mvvm |  · 9W 15T 0L (9.2↑8.8) |  · 5W 17T 1L |  · 21W 0T 0L (9.1↑6.2) |
| uikit-mvvm |  · 20W 2T 2L (9.1↑8.2) |  · 24W 0T 0L (9.0↑7.4) |  · 24W 0T 0L (9.2↑5.7) |
| gcd-operations |  · 15W 9T 0L (8.6↑8.1) |  · 18W 1T 5L (8.6↑7.8) |  · 12W 0T 1L (8.8↑7.1) |
| ios-testing |  · 23W 7T 0L (8.9↑8.0) |  · 6W 21T 0L |  · 27W 0T 0L (9.2↑6.5) |
| swift-concurrency |  · 15W 9T 0L (8.9↑8.5) |  · 21W 0T 0L (9.0↑7.6) |  · 21W 0T 0L (9.2↑7.3) |
| ios-security |  · 9W 15T 0L (9.3↑8.9) | · 2W 15T 0L |  · 10W 0T 0L (9.1↑6.0) |
| tca-swiftui |  · 14W 2T 4L (8.7↑7.8) | — | — |
| viper-uikit |  · 15W 1T 0L (9.5↑9.4) |  · 16W 0T 0L (9.0↑7.3) |  · 16W 0T 0L (7.3↑1.7) |
| ios-logging |  · 23W 3T 1L (9.0↑8.5) |  · 23W 0T 4L (8.8↑8.2) |  · 24W 0T 0L (9.2↑6.3) |
| ios-accessibility |  · 27W 5T 4L (8.5↑7.9) |  · 31W 1T 4L (9.2↑8.5) |  · 36W 0T 0L (9.0↑7.2) |
| ios-localization |  · 10W 21T 5L (8.9↑8.7) |  · 26W 2T 8L (9.1↑8.9) |  · 36W 0T 0L (9.0↑6.3) |
Each cell: assertion delta (with skill vs without) · A/B blind comparison W/T/L with avg scores where available. Colored badge = best result for that skill. "—" = not yet benchmarked.
See BENCHMARKING.md for full methodology.
Install
Claude Code Plugin (recommended)
claude plugin add rusel95/ios-agent-skillsInstall individual skills:
claude plugin add rusel95/ios-agent-skills --skill swiftui-mvvm
claude plugin add rusel95/ios-agent-skills --skill swift-concurrencyAgent Skills CLI
npx skills add rusel95/ios-agent-skills --skill swiftui-mvvmManual
Clone and copy the skills/ directory into your project.
What Makes These Different
- Production-first — every pattern comes from real enterprise codebases, not tutorials
- Iterative refactoring — small, reviewable PRs (≤200 lines) instead of "rewrite everything" approaches
- Anti-pattern prevention — AI tools consistently generate broken patterns (retain cycles in VIPER, outdated TCA APIs, unsafe GCD). These skills prevent that
- Architecture coverage — the only collection covering VIPER, TCA, GCD, Security Audit, Accessibility, and Localization. No other iOS skill marketplace covers these domains
- Rigorously benchmarked — the most comprehensively evaluated iOS skill collection available: 850+ discriminating assertions across 260+ scenarios, tested against 3 LLMs with blind A/B quality scoring. Every skill ships with reproducible eval data
Author
Ruslan Popesku — Lead iOS Software Engineer GitHub · LinkedIn
License
MIT