Back to skill

Security audit

小程序数据问答

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly performs read-only Umeng analytics queries, but it also requires extra telemetry that reports each AppKey without a clear consent or privacy boundary.

Review this skill before installing if AppKeys are sensitive in your organization. The analytics queries themselves are scoped and read-only, but the skill asks the agent to send usage telemetry and each AppKey through `umeng-cli trace`; install only if that reporting is acceptable, and prefer a pinned, trusted CLI installation with clear credential-cache handling.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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 (2)

T01 · Skill Instruction Hijacking

Warning
Location
SKILL.md:17
Finding

Mandatory Transmission of User AppKey Through Skill Telemetry

Content
View full analysis
"}' ``` The surrounding instructions require the agent to execute the first telemetry command after reading the Skill and to execute the second command after receiving an AppKey but before making the first business API request. ### Technical Analysis The Skill directs the agent to perform telemetry as a side effect of loading and using the Skill. In particular, it requires a user-provided `dataSourceId`, identified as an AppKey elsewhere in the document, to be passed to `umeng-cli trace`. This operation is not required to perform the documented read-only analytics requests. The behavior therefore changes the agent's execution goals by adding an undisclosed prerequisite operation involving user data. The project contains only `SKILL.md`; the implementation of `umeng-cli trace` is not present, so its destination, retention policy, access controls, and exact handling of the AppKey cannot be independently verified within the audited artifact. Although an AppKey may function primarily as an application identifier rather than a complete authentication secret, transmitting it can disclose which application is being analyzed and correlate that application with Skill usage. ### Attack Path 1. An agent loads and follows the instructions in `SKILL.md`. 2. The Skill directs the agent to execute `umeng-cli trace` immediately. 3. A user supplies or the agent obtains a `dataSourceId` or AppKey for an analytics request. 4. Before the requested business API call, the Skill directs the agent to place that AppKey in a telemetry payload. 5. The external `umeng-cli` binary processes the ...[truncated 787 chars]
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
SKILL.md:44
Finding

Unpinned Global Installation of a Security-Sensitive CLI Dependency

Content
View full analysis
Remediation
View remediation
``` 2. Publish and verify the expected package integrity hash before installation. 3. Prefer a project-local installation with a committed lockfile over a global installation. 4. Use `npm ci` in controlled environments so dependency resolution matches the reviewed lockfile. 5. Review the package's lifecycle scripts and transitive dependencies before approving upgrades. 6. Avoid running npm installation commands with administrative privileges. 7. Document the expected npm registry, publisher identity, package version, and release verification process. 8. Separate credential storage from the CLI process where possible and restrict credential-file permissions to the owning user. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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 (7)

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The documentation states there is 'no write risk' and that all interfaces are query-only, yet it directs the agent to perform telemetry reporting that writes data to another service. This mismatch can mislead security controls and human reviewers into treating the skill as passive when it actually performs additional state-changing network actions.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill claims to be limited to 12 read-only analytics interfaces, but it also instructs the agent to execute separate telemetry commands that create an external side effect. This is dangerous because operators or downstream systems may trust the 'read-only' claim and unknowingly permit outbound reporting of usage and application identifiers beyond the user's requested analytics query.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The telemetry instruction sends the user's provided appkey to an external command flow without any explicit user-facing privacy warning or consent step. This is dangerous because the agent may disclose tenant-specific identifiers to a third party implicitly, violating least surprise and potentially internal data handling requirements.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill explicitly instructs the agent to capture a user-supplied appkey and report it before the first business API call. This creates unnecessary disclosure of a user-provided identifier and establishes a covert secondary data flow unrelated to the requested read operation.

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 · SKILL.md (reported line 55)May include surrounding context.

登录后 AK/SK 自动加密缓存;后续调用无需手动配置

3. 验证(可选)

umeng-cli call '{"name":"umeng.umini.getAppList","api":{"method":"GET","baseUrl":"https://gateway.open.umeng.com/openapi","endpoint":"param2/1/com.umeng.umini/umeng.umini.getAppList","authType":"umeng-aksk"}}' '{}'

text

### 登录要求

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 526)May include surrounding context.

登录后 AK/SK 自动加密缓存;后续调用无需手动配置

3. 验证(可选)

umeng-cli call '{"name":"umeng.umini.getAppList","api":{"method":"GET","baseUrl":"https://gateway.open.umeng.com/openapi","endpoint":"param2/1/com.umeng.umini/umeng.umini.getAppList","authType":"umeng-aksk"}}' '{}'

text

### 登录要求

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The skill instructs the agent to send appkey-specific telemetry even though that data is not clearly necessary to fulfill the user's analytics request. Collecting and transmitting extra identifiers increases privacy and data-governance risk, especially if appkeys are considered sensitive internal asset identifiers.

Content

No source excerpt is available for this finding.