Back to skill

Security audit

ios-self-improve

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly an iOS code-checking helper, but its rule-proposal command runs an unpinned dependency script while some security docs describe the skill as read-only with no code execution.

Review and trust developer-self-improve-core before installing this skill, because using propose runs that dependency's shell script. Prefer a pinned or otherwise verified dependency version, and treat the current read-only/no-code-execution claims as incomplete.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (1)

T08 · Insecure Dependencies

Warning
Location
scripts/ios-self-improve.sh:258
Finding

Unpinned Dependency Executed Without Authenticity or Integrity Validation

Content
View full analysis

Vulnerability Details

File Location: scripts/ios-self-improve.sh:258-263
Related Locations: .clawhub.json:11-13, SKILL.md:27-32, README.md:9-16
Vulnerability Type: Insecure dependency execution and supply-chain trust
Risk Level: Medium

Vulnerable Code

bash
# Call developer-self-improve-core's post-check
local auto_script="$SKILL_DIR/../developer-self-improve-core/scripts/developer-self-improve-core.sh"

if [ -f "$auto_script" ]; then
    "$auto_script" post-check "$content" "$scene"
else
    echo "⚠️  developer-self-improve-core is not installed; unable to generate a draft"
fi

The dependency is declared without a version or immutable identifier:

json
"dependencies": [
  "developer-self-improve-core"
]

The installation instructions likewise request an unversioned package:

bash
clawhub install developer-self-improve-core

Technical Analysis

The propose operation executes a shell script from a predictable sibling directory. The only runtime validation is Bash's -f test, which establishes that the path is a regular file but does not verify:

  • Package version or immutable digest
  • Cryptographic signature or checksum
  • File ownership and permissions
  • Canonical path and package identity
  • Whether the sibling installation has been replaced or modified
  • Whether the dependency release is the version previously audited

Consequently, the effective code executed by this Skill is not fully represented by the audited package. A compromised dependency release, package-resolution substitution, or attacker-controlled sibling directory can alter runtime behavior.

This behavior also conflicts with security documentation claiming that the Skill does not execute dependency or external code. The dependency is a separately installed component, and its script is directly invoked.

The arguments are passed using proper shell quoting, so no direct shell-command injection through $content or $scene was identifie ...[truncated 1734 chars]

Remediation
View remediation

Remediation Suggestions

  1. Pin developer-self-improve-core to an audited immutable version or content digest in package metadata and installation documentation.
  2. Require cryptographic package-signature or checksum verification before the dependency is installed or executed.
  3. Resolve the dependency through a trusted package-manager API instead of assuming that a writable sibling path represents the declared package.
  4. Before execution, canonicalize the path and verify that it remains within the expected dependency installation directory.
  5. Validate file ownership and reject scripts writable by unexpected users or groups.
  6. Replace direct shell-script invocation with a constrained, versioned interface that exchanges validated structured data where feasible.
  7. Document accurately that the propose operation executes dependency code and inherits that dependency's security properties.
  8. Audit the pinned dependency together with this Skill, particularly its handling of the user-controlled content and scene arguments.
  9. Correct the related inaccurate claims that all operations are read-only, that no dependency code is executed, and that file existence constitutes dependency validation.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (25)

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

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

md
## Security Measures

- ✅ No dangerous commands (rm -rf, curl | sh, eval, etc.)
- ✅ No external network access
- ✅ No system-level permissions
- ✅ All operations are logged

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill description is written as a Chinese-only user-facing instruction set, and there is no indication anywhere in the README that users may choose another language or that the locale restriction is required for a region-specific purpose. Under the policy, forcing a specific language without opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

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 · README.md (reported line 32)May include surrounding context.

md
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 117)May include surrounding context.

md
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 152)May include surrounding context.

md
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 160)May include surrounding context.

md
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 303)May include surrounding context.

md
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 308)May include surrounding context.

md
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 · SUBMISSION_NOTE.md (reported line 35)May include surrounding context.

md
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 · config/config.yaml (reported line 51)May include surrounding context.

yaml
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 · config/config.yaml (reported line 53)May include surrounding context.

yaml
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 · config/config.yaml (reported line 54)May include surrounding context.

yaml
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 · config/config.yaml (reported line 55)May include surrounding context.

yaml
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 · scripts/ios-self-improve.sh (reported line 40)May include surrounding context.

sh
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 · scripts/ios-self-improve.sh (reported line 92)May include surrounding context.

sh
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 · scripts/ios-self-improve.sh (reported line 93)May include surrounding context.

sh
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 · scripts/ios-self-improve.sh (reported line 197)May include surrounding context.

sh
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

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 · scripts/ios-self-improve.sh (reported line 198)May include surrounding context.

sh
- ✅ 崩溃风险检测(数组越界、空指针、循环引用)
- ✅ 导航栏配置检查
- ✅ Info.plist 隐私权限检查
- ✅ AutoLayout 约束检查
- ✅ SwiftUI 生命周期检查
- ✅ 异步代码检查

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
85% confidence
Finding

The description is entirely in Chinese and frames the skill as a Chinese-language iOS development skill, with no indication that users may choose another language. This creates a natural-language locale constraint without documented opt-in or justification as a region-specific tool.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Describing the dependency relationship as 'read-only access' is inconsistent with later admitting script execution capability, which goes beyond passive reads and can trigger arbitrary behavior within the dependency's exposed scripts. This weakens the stated trust boundary and may mislead operators into granting the skill more trust than its actual behavior warrants.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The trust model makes a materially misleading security claim: it says there is 'no execution of external code' while also stating the skill can call dependency scripts for rule generation. Executing scripts from a dependency is code execution across a trust boundary, and understating that capability can cause users or reviewers to underestimate supply-chain and runtime risk.

Content

No source excerpt is available for this finding.

Unrestricted Tool Access

Medium
Category
Excessive Agency
Confidence
80% confidence
Finding

Skill grants unrestricted tool access without appropriate constraints. An agent with unfettered tool access can perform arbitrary actions including file modification, network requests, and code execution.

Content

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

md
**This skill CANNOT:**
- ❌ Modify `developer-self-improve-core` files
- ❌ Execute arbitrary code from the dependency
- ❌ Access system-level resources
- ❌ Make external network calls

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The entire skill file is presented only in Chinese, including the title and all rule descriptions, with no indication that the user can choose another language or that the skill is intentionally restricted to a Chinese-speaking audience. This creates a natural-language locale policy concern because it implicitly enforces a specific language without opt-in or documented justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

This manifest/config file contains user-facing instructional text entirely in Chinese, including setup guidance and examples. Under the language/locale policy, forcing a specific language without offering user choice or documenting a justified locale restriction can be a policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

This shell script's user-facing comments, help text, examples, and runtime messages are all presented in Chinese, including the primary usage guidance shown to users. Because the file does not offer any language or locale selection or document a justified region-specific constraint, it creates a locale policy issue under the natural-language policy rule.

Content

No source excerpt is available for this finding.