agentleFS
apexbymanishapexbymanish/claude-ai-agents-ios/.claude/skills/ios-security-review/SKILL.md

ios-security-review

Skill0 starsChanged 3 months agoMITView on GitHub

Comprehensive security audit procedure for iOS/Swift code covering data storage, transport security, authentication, input validation, deep links, third-party SDKs, and entitlements. Use when asked for a security review, "is this secure", to check for vulnerabilities, or to audit authentication/authorization.

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If the app implements certificate or public-key pinning, that this feature's networking path doesn't route around it.

    line 58
  2. Distinctive:

    Every finding names the file:line evidence and is phrased as something to verify (e.g.

    line 187

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

iOS Security Review

Per ios-evidence-reporting's "a pattern match is a lead, not a finding" rule, every hit below is something to check, not a proven exploit. This procedure is read-only reconnaissance; it produces findings, not fixes.

1. Data storage & privacy

grep -rn "UserDefaults" --include="*.swift" . | grep -iE 'token|password|secret|credential|auth'
grep -rl "kSecClass\|SecItemAdd\|SecItemCopyMatching" --include="*.swift" . | wc -l   # Keychain usage at all?
grep -rn "NSFileProtectionType\|isExcludedFromBackup" --include="*.swift" .
grep -rn "UIPasteboard.general" --include="*.swift" .

Check for:

  • Credentials/tokens/PII in UserDefaults, a plist, or unencrypted Core Data/SwiftData instead of the Keychain.
  • Sensitive Keychain items using an appropriate accessibility class (e.g. kSecAttrAccessibleWhenUnlockedThisDeviceOnly for high-sensitivity secrets, not the default).
  • Sensitive files not excluded from iCloud/iTunes backup when they should be (isExcludedFromBackup).
  • Sensitive data placed on the general UIPasteboard without an expiration or local-only option — it can be read by other apps.
  • Sensitive screens (auth, payment) not redacted in the app switcher snapshot (no handling of UIApplication.willResignActiveNotification to blank the screen before backgrounding).

Known false positives: a UserDefaults key whose name coincidentally contains "token"/"password" but stores a non-sensitive flag (e.g. hasSeenTokenOnboarding, passwordFieldFocused as UI state) rather than a credential value — read what's actually assigned, not just the key name, before flagging it.

2. Transport security

grep -A5 "NSAppTransportSecurity" **/*.plist 2>/dev/null
grep -rn "NSAllowsArbitraryLoads\|NSExceptionAllowsInsecureHTTPLoads" **/*.plist 2>/dev/null
grep -rn "didReceive challenge" --include="*.swift" . -A8 | grep -i "useCredential\|performDefaultHandling"
grep -rn "http://" --include="*.swift" . | grep -v "localhost\|127.0.0.1"

Check for:

  • ATS weakened globally (NSAllowsArbitraryLoads: true) rather than a scoped, justified per-domain exception.
  • TLS certificate/server-trust validation bypassed (URLSession(didReceive:completionHandler:) unconditionally calling .useCredential for any host) — this defeats TLS entirely, not just weakens it.
  • If the app implements certificate or public-key pinning, that this feature's networking path doesn't route around it.
  • Sensitive data (tokens, PII) placed in a URL query string instead of a header or POST body — query strings end up in server logs, proxy logs, and browser/webview history.

Known false positives: NSAllowsArbitraryLoads set to false at the top level with a scoped NSExceptionDomains entry for one known test/staging host is a justified, narrow exception, not a blanket weakening — read the surrounding plist keys, not just whether the string appears. A .useCredential response inside didReceive challenge for NSURLAuthenticationMethodClientCertificate (presenting a client certificate for mutual TLS) is not the same finding as .useCredential for NSURLAuthenticationMethodServerTrust (bypassing server validation) — check which authentication method the challenge handler is actually branching on before treating it as a TLS bypass.

3. Authentication & session management

grep -rln "LAContext\|LocalAuthentication" --include="*.swift" .
grep -rn "func logout\|func signOut" --include="*.swift" . -A15

Check for:

  • Biometric auth (LAContext/Face ID/Touch ID) gating access to an actual secret (a Keychain item with an access-control policy) rather than just flipping an app-level boolean that any code path could bypass — a boolean gate protects nothing if the underlying data is reachable without it.
  • Token/session expiry and refresh actually enforced, not stored and reused indefinitely.
  • Logout actually clears stored tokens/session state (Keychain, in-memory, cookies) rather than only navigating to a login screen while the old session remains valid underneath.

4. Input validation & injection

grep -rn "WKWebView\|UIWebView" --include="*.swift" .
grep -rln "WKScriptMessageHandler" --include="*.swift" .
grep -rn "URL(string:.*\\\\(" --include="*.swift" .   # string-interpolated URLs

Check for:

  • UIWebView (deprecated, no process isolation) instead of WKWebView.
  • A WebView loading arbitrary or user-supplied URLs without a scheme/host allowlist.
  • A JS-to-native bridge (WKScriptMessageHandler) exposing more native functionality than the loaded content needs.
  • URLs or query strings built by string interpolation/concatenation of unescaped user input rather than URLComponents.
  • Deserialized JSON/response data used without validating shape/types before acting on it — a malformed or unexpected response shouldn't corrupt app state or crash.

5. Deep links & URL schemes

grep -rn "CFBundleURLSchemes\|associatedDomains" **/*.plist **/*.entitlements 2>/dev/null
grep -rln "func application.*openURL\|onOpenURL" --include="*.swift" .

Check for:

  • A custom URL scheme used for anything sensitive — any app can register the same scheme and intercept it; prefer Universal Links/associated domains for anything security-relevant.
  • Incoming URL parameters used to trigger navigation or actions without validation — treat a deep link parameter with the same suspicion as network input, not as trusted internal state.
  • Any sensitive action (login, payment, account change) reachable directly via an unauthenticated deep link.

6. Third-party SDKs & dependencies

find . -name "Podfile.lock" -o -name "Package.resolved"
grep -rln "Analytics\|Crashlytics\|FirebaseCrashlytics" --include="*.swift" .

Check for:

  • Data handed to analytics/crash-reporting SDKs that includes PII, tokens, or other secrets (check what's passed to logging/tracking calls near the feature you're reviewing, not just that the SDK exists).
  • Dependencies pinned to versions with known vulnerabilities, if a vulnerability database check is available in this environment.

7. Code-level hygiene

grep -rniE '(api[_-]?key|secret|password|token)\s*=\s*"' --include="*.swift" --include="*.m" --include="*.h" .
grep -rniE '(print|NSLog|os_log)\(.*\b(token|password|secret)\b' --include="*.swift" .

Check for:

  • Hardcoded API keys/secrets in source rather than build-time injected config or a server-side proxy.
  • Debug logging left in that would print request/response bodies, tokens, or PII — check whether it's gated out of release builds.

Known false positives: a match inside a Tests/UITests target using an obviously fake value ("test-api-key-123", "fake-token") is a test fixture, not a production secret — check the target membership before flagging. An enum case declaration (case token, case secret) is a type discriminator with no associated value, not an assigned credential; only a = "..." assignment is a hit worth reporting.

Jailbreak/tamper detection is a judgment call, not a default recommendation: proportionate for an app handling high-value secrets (banking, enterprise MDM-managed data), unnecessary overhead for most apps. Say so explicitly rather than recommending it reflexively.

8. Entitlements & capabilities

find . -name "*.entitlements" -exec cat {} \;

Check for:

  • Keychain access groups or App Groups shared more broadly than the feature actually requires.
  • Requested capabilities/entitlements that don't match anything the code visibly uses — unused broad permissions are attack surface with no corresponding benefit.

Reporting

Every finding names the file:line evidence and is phrased as something to verify (e.g. "3 UserDefaults calls near token in AuthManager.swift:42-58 — verify these aren't the actual session token, not just a non-sensitive flag"), never as a confirmed exploit unless you've actually traced the data flow and confirmed it. Close with the ios-evidence-reporting skill's status block (e.g. STORAGE, TRANSPORT, AUTH, INPUT-VALIDATION, DEEP-LINKS, DEPENDENCIES, CODE-HYGIENE, ENTITLEMENTS) — include only the categories that were actually relevant to what you reviewed.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.