Back to skill

Security audit

技能更新管理器

Security checks for vulnerabilities and agentic risk

Overview

This is a coherent skill updater, but it can persistently change agent memory and contact or fetch updates from insufficiently constrained remote URLs.

Install only if you are comfortable with a skill that manages all installed-skill update records, writes to shared MEMORY.md, and performs outbound update checks. Keep update checks disabled unless needed, review every update URL before checking, avoid registering untrusted ClawHub URLs, and do not let it clone or download updates unless you independently trust the source and can verify the exact version being installed.

Vulnerability Patterns
  • 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
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (4)

T02 · Agent Memory Poisoning

Error
Location
scripts/skill_registry.py:32
Finding

Automatic Injection of Persistent Instructions into Shared Agent Memory

Content
View full analysis
--source local # 更新配置 python skills/skills-update-manager/scripts/skill_registry.py --action update_config --enable-updates true ``` """ ...[truncated 2282 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
scripts/skill_registry.py:413
Finding

Arbitrary URL Requests Through Unvalidated ClawHub Update Addresses

Content
View full analysis
Optional[str]: """ 从ClawHub获取最新版本号 Args: skill_url: ClawHub技能页面URL Returns: 最新版本号 """ try: response = requests.get(skill_url, timeout=10) ``` ### Technical Analysis The `clawhub` source value is treated as if it guarantees that `update_url` belongs to ClawHub, but the implementation never enforces that relationship. An attacker or untrusted registry entry can therefore supply URLs such as loopback addresses, private-network hosts, link-local cloud metadata addresses, or attacker-controlled redirectors. ...[truncated 1596 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/skill_registry.py:469
Finding

Global Substring Matching Can Unexpectedly Enable Network Update Checks

Content
View full analysis
Remediation
View remediation

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:159
Finding

Remote Skill Updates Are Retrieved Without Integrity or Authenticity Verification

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (16)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 57)May include surrounding context.

md
2. 读取 `SKILL.md` 文件

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 64)May include surrounding context.

md
2. 读取 `SKILL.md` 文件

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
95% confidence
Finding

The skill describes capabilities that read and write local files and perform network access, but it declares no explicit tool scope or permission boundaries. That creates an over-privileged, under-specified execution model where an agent may perform sensitive actions without clear user visibility or enforcement constraints.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The manifest description and the entire operational guidance are written in Chinese, with no indication that the skill supports other languages or that Chinese is a justified region-specific requirement. Under the policy, language constraints should be opt-in or explicitly documented as necessary.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
88% confidence
Finding

The skill sends outbound requests to GitHub and other update sources based on registered skill metadata, which can disclose installed-skill inventory and trigger untrusted remote content retrieval workflows. In context, this becomes more dangerous because the skill also instructs the agent to use git clone or download-based updates later, creating a supply-chain path from remote metadata to local installation activity.

Content

Scanner excerpt · SKILL.md (reported line 141)May include surrounding context.

md
1. 读取 `MEMORY.md` 配置,判断更新检查是否启用
2. 如果启用,遍历所有已注册技能
3. 根据技能来源类型调用对应的版本检查逻辑:
   - **GitHub**:访问 `https://api.github.com/repos/{owner}/{repo}/releases/latest`
   - **ClawHub**:访问技能页面,解析 "Current version" 字段
4. 对比本地版本与远程版本
5. 返回有更新的技能列表

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/registry_format.md (reported line 84)May include surrounding context.

md
### GitHub (github)
- 标识符:`github`
- 版本获取方式:访问 GitHub API
- API端点:`https://api.github.com/repos/{owner}/{repo}/releases/latest`
- 返回字段:`tag_name` 或 `name`

**注意事项**:

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

User-facing descriptions, messages, and CLI help text are written in Chinese throughout the file, with no option for users to select another language. This can constitute a locale/language policy violation when a skill enforces a specific language without user opt-in or documented regional scope.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The script reaches outside its own registry scope and modifies a global MEMORY.md file under a workspace path. In an agent environment, altering shared memory/state can influence other components, create persistence, and change behavior without clear user consent, which is a genuine integrity and scope-boundary issue.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The constructor automatically initializes and may modify MEMORY.md on object creation, without an explicit action dedicated to shared-memory mutation or a user-facing warning. In a skill system, silent state changes are dangerous because merely loading the component can create persistent side effects.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The docstring states the operation is append-only, but the implementation rewrites the entire MEMORY.md file with mode 'w'. This mismatch is dangerous because users and reviewers may trust a less invasive behavior while the code can overwrite or corrupt shared memory contents if the in-memory content is stale or manipulated.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
87% confidence
Finding

The code transmits data to an external service (GitHub API) as part of update checking. In context this is functional rather than overtly malicious, but it is still security-relevant because it creates network egress from the agent environment and may expose information about installed skills and usage.

Content

Scanner excerpt · scripts/skill_registry.py (reported line 394)May include surrounding context.

python
return None
            
            owner, repo = match.groups()
            api_url = f"https://api.github.com/repos/{owner}/{repo}/releases/latest"
            
            response = requests.get(api_url, timeout=10)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The update check performs outbound HTTP requests to GitHub without explicit runtime disclosure. This can leak installed skill names/repository identifiers and network metadata, and in restricted environments any undeclared egress is a meaningful security/privacy issue.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The ClawHub version check fetches arbitrary skill URLs without explicit user-facing notice. Because update_url can be stored per skill, this creates undeclared egress and may contact untrusted domains, increasing privacy, SSRF-like, and policy-compliance risk in agent deployments.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The update_config path is documented as append-only but rewrites MEMORY.md in place when existing config is found. Rewriting a shared file increases risk of unintended modification, lost concurrent changes, and hidden persistence in a broader agent workspace.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

该文件标题及全文说明均以中文呈现,且没有声明可使用其他语言、提供本地化选项,或说明这是仅面向中文用户的区域性规范。根据语言/区域政策要求,未经用户选择而固定单一语言属于自然语言层面的政策问题。

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The _save_registry() method writes skills_registry.json directly, but there is no visible log, print, or confirmation when persisting changes. While file writes are expected for a registry manager, this specific operation lacks any explicit user disclosure in code output.

Content

No source excerpt is available for this finding.