Back to skill

Security audit

mobile-release

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marked this skill benign because its findings are treated as expected in context, though one scanner reports that release uploads and submissions exceed the stated readiness and planning purpose; confidence is medium and coverage is incomplete.

The verdict is benign, not risk-free. Before installing, consider whether you want a skill whose reported capabilities include uploading a build to Google Play production and submitting an iOS build for review. Scanner coverage was partial: obfuscated instruction text was not fully evaluated and several static checks were degraded.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (9)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The description presents a readiness-checking and planning skill, but the supplied calls perform release-changing operations. Although rollout strategy is within the declared subject matter, executing uploads and review submissions exceeds checking or advising. The BAD/GOOD comments distinguish deployment approaches but do not disable the unconditional calls; if executed as supplied, both Google Play upload calls would run.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 51)May include surrounding context.

md
nifests | Since 12 February 2025, submissions must contain a valid privacy manifest for listed commonly used SDKs; binary SDKs on the list also need a signature | Apple "Adding a privacy manifest to your app or third-party SDK"; "Third-party SDK requirements" |
| App Review 4.8 Login Services | Third-party or social login for the primary account requires an equivalent login option that limits data to name and email, lets the user keep the email private, and does not collect interactions for ads without consent. Listed exceptions apply | App Store Review Guidelines 4.8 |
| App Review 5.1.1(v) | Apps that support account creation must offer account deletion within the app | App Store Review Guidelines 5.1.1(v) |
| App Store phased release | 7 days: 1%, 2%, 5%, 10%, 20%, 50%, 100% of users with automatic updates. Pause for up to 30 days in total. Anyone can still download the update manually | App Store Connect Help "Release a version update in phases" |
| TestFlight | Up to 100 internal

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · references/app-store.md (reported line 128)May include surrounding context.

md
, login or splash screen | Screenshot set per device size |
| 2.5.4 | Background services only for intended purposes (VoIP, audio, location, task completion, notifications) | `UIBackgroundModes` justified, BLE background modes used for real features |
| 4.2 Minimum Functionality | More than a repackaged website | WebView-only apps |
| 4.8 Login Services | Third-party or social login for the primary account needs an equivalent option limiting data to name and email, private email, no ad tracking without consent | Sign in with Apple or another compliant option, or a listed exception |
| 5.1.1(i) Privacy Policies | Privacy policy link in App Store Connect metadata and easily accessible in the app | Link in settings or about screen |
| 5.1.1(ii) Permission | Purpose strings clearly and completely describe the use | Every `NS*UsageDescription` specific to the feature |
| 5.1.1(iv) Access | No manipulating or forcing consent; offer alternatives when a permission is declined | App works with

Session Persistence

Medium
Category
Rogue Agent
Confidence
75% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 62)May include surrounding context.

md
number on each store, target date, and who approves. Ask; do not guess store tracks or groups.
2. **Read the build configuration.** Android: `targetSdk`, `minSdk`, `versionCode`, `versionName`,
   signing config, merged manifest permissions and services. iOS: deployment target, Xcode version in CI,
   `CFBundleShortVersionString`, `CFBundleVersion`, `Info.plist` purpose strings, entitlements,
   `PrivacyInfo.xcprivacy`. Flutter: `version:` in `pubspec.yaml`, plus both native projects.
3. **Inventory SDKs and data.** From `libs.versions.toml`, Gradle dependencies, `Package.resolved`,
   `Podfile.lock`, `pubspec.lock`. For each SDK: what it collects, whether it tracks, its privacy manifest.

Session Persistence

Medium
Category
Rogue Agent
Confidence
75% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 271)May include surrounding context.

text

```xml
<!-- Info.plist: required when requesting authorization; explain the actual use -->
<key>NSUserTrackingUsageDescription</key>
<string>Allows ads based on your activity in other companies' apps and websites.</string>

Session Persistence

Medium
Category
Rogue Agent
Confidence
75% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/app-store.md (reported line 114)May include surrounding context.

md
is all zeros; tracking with another identifier (for example a hashed email) is not allowed.
- Not tracking: on-device-only linking, sharing solely for fraud prevention or security, consumer reporting
  for credit decisions.
- `NSUserTrackingUsageDescription` in `Info.plist` explains the use. No incentives and no gating (5.1.2(i)).
- Fingerprinting is not allowed regardless of ATT.

## Common App Review rejections

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
70% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · SKILL.md (reported line 107)May include surrounding context.

yaml
# BAD: CI pinned to an old Xcode; App Store Connect rejects the upload
- run: sudo xcode-select -s /Applications/Xcode_16.4.app

# GOOD: Xcode 26 or later (exact path depends on the runner image), checked in the job log
- run: sudo xcode-select -s /Applications/Xcode_26.0.app

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
70% confidence
Finding

Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.

Content

Scanner excerpt · SKILL.md (reported line 110)May include surrounding context.

  • run: sudo xcode-select -s /Applications/Xcode_16.4.app

GOOD: Xcode 26 or later (exact path depends on the runner image), checked in the job log

  • run: sudo xcode-select -s /Applications/Xcode_26.0.app
  • run: xcodebuild -version
text

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 340)May include surrounding context.

md
## Anti-patterns

- Declaring a build "store ready" without checking the merged manifest and the archived bundle.
- Copying Data safety or App Privacy answers from another app, or answering "no data collected" with
  crash reporting or analytics SDKs in the build.
- Requesting `ACCESS_BACKGROUND_LOCATION` for BLE scanning that does not derive location.