Back to skill

Security audit

未来清晨

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent event lookup helper, but its detail lookup asks the agent to run inline JavaScript with an unvalidated activity ID, which could execute unintended code if a malicious ID is supplied.

Review before installing. Use this only for explicit 未来清晨 event queries, and do not let arbitrary user-supplied activity IDs be pasted into the inline Node.js command. Prefer a revised version that validates campaign IDs as UUIDs and passes them as data, not as source code inside node -e.

Vulnerability Patterns
  • 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
  • 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)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:91
Finding

Arbitrary JavaScript Execution Through Unsafe Campaign ID Substitution

Content
View full analysis
{ ``` The complete command is executed as an inline, double-quoted `node -e` shell command. The instructions require replacing the placeholder inside a single-quoted JavaScript string with the activity ID supplied for the query. ### Technical Analysis The campaign ID is inserted directly into executable JavaScript source without escaping or validating it. Consequently, an attacker-controlled ID containing a single quote can terminate the `CAMPAIGN_ID` string literal. The attacker can then append arbitrary JavaScript statements that Node.js executes with the privileges of the process invoking the Skill. For example, a value structurally similar to the following could close the original string, execute attacker-selected Node.js code, and restore valid syntax: ```text '; ATTACKER_CONTROLLED_JAVASCRIPT; const placeholder=' ``` This is a source-code injection vulnerability rather than an API-level issue. JSON serialization inside `post()` does not mitigate it because execution occurs while Node.js parses the generated source, before the HTTP request is constructed. The risk is especially significant because Node.js provides direct access to the filesystem, environment variables, networking, and child-process APIs. ### Attack Path 1. An attacker asks the Agent to retrieve details for a maliciously crafted activity ID. 2. Following line 91, the Agent replaces the `CAMPAIGN_ID` placeholder in the inline command with the supplied value. 3. The malicious value terminates the JavaScript string literal at line 97 and introduces additional JavaScript statements. 4. The generated command is passe ...[truncated 1195 chars]
Remediation
View remediation
{ // Process the response. }); ``` 3. **Avoid shell-based execution** - Use an execution facility equivalent to `spawn` or `execFile` with separate arguments. - Do not construct commands through string concatenation. - Do not rely solely on quote escaping, as layered shell and JavaScript parsing makes escaping fragile. 4. **Reduce process privileges** - Run the Skill with the minimum filesystem, environment, process, and network permissions required for activity queries. - Do not expose unrelated credentials to the process. - Where supported, restrict outbound network access to the documented API host. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (5)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger condition is broad enough that any mention of “未来清晨” can invoke the skill, even when the user is not requesting event data. Over-broad activation can cause unintended external API calls and disclosure of user context to the third-party service, especially in multi-skill environments where routing is based on loose matching.

Content

No source excerpt is available for this finding.

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 · SKILL.md (reported line 20)May include surrounding context.

text
CLUB_ID = 018ce711-aeab-f611-7900-486c089013e4
API_BASE = https://api.cumen.fun/api/xx.cumen.v1.CumenService

功能一:列出所有活动

External Transmission

Medium
Category
Data Exfiltration
Confidence
79% confidence
Finding

This code performs a live POST request to a third-party endpoint whenever the skill lists events. In the context of an agent skill, external transmission is security-relevant because invocation may occur automatically from broad triggers, sending request metadata and potentially exposing user-derived context to an external service without explicit consent boundaries.

Content

Scanner excerpt · SKILL.md (reported line 28)May include surrounding context.

bash
node -e "
const https = require('https');
const API = 'https://api.cumen.fun/api/xx.cumen.v1.CumenService/ListCampaignsOfClub';
const CLUB_ID = '018ce711-aeab-f611-7900-486c089013e4';

function post(url, body) {

External Transmission

Medium
Category
Data Exfiltration
Confidence
79% confidence
Finding

This code sends campaign IDs to an external API to retrieve details, which is expected functionally but still represents third-party data transmission. In an agent setting, this becomes a true security concern because untrusted or user-supplied IDs may be forwarded externally, and the skill has no documented consent, validation, or privacy boundary around these outbound requests.

Content

Scanner excerpt · SKILL.md (reported line 96)May include surrounding context.

bash
node -e "
const https = require('https');
const API = 'https://api.cumen.fun/api/xx.cumen.v1.CumenService/GetCampaign';
const CAMPAIGN_ID = 'CAMPAIGN_ID';

function post(url, body) {

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
79% confidence
Finding

The file is written to operate and present outputs in Chinese, including the described default/output formatting behavior, but it does not offer users a language or locale choice. Under the policy, forcing a specific language without opt-in can be a natural-language policy violation unless the locale restriction is explicitly justified.

Content

No source excerpt is available for this finding.