Back to skill

Security audit

aigc-web-push · AI内容生成与全网推送

Security checks for vulnerabilities and agentic risk

Overview

The skill is marked suspicious because scanners report default automatic replacement of installed files and a bulk draft deletion capability; analysis was partial, so the evidence is incomplete.

Consider avoiding installation unless you are comfortable with automatic updates that replace local skill files and the reported draft deletion capability. Keep in mind that only 8 of 12 components were fully inspected; four were partial, including the push scripts, so the scan does not establish that all behavior is safe.

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
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
Findings (25)

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
80% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · Humanizer.md (reported line 170)May include surrounding context.

md
**After:**
> The term is primarily promoted by Dutch institutions, not by the people themselves. You don't say "Netherlands, Europe" as an address, yet this mislabeling continues in official documents.
**Before:**
> The new policy — announced without warning — affects thousands of workers. The changes -- long overdue according to critics -- will take effect immediately.
**After:**
> The new policy, announced without warning, affects thousands of workers. The changes, long overdue according to critics, will take effect immediately.

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
80% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · Humanizer.md (reported line 172)May include surrounding context.

md
**Before:**
> The new policy — announced without warning — affects thousands of workers. The changes -- long overdue according to critics -- will take effect immediately.
**After:**
> The new policy, announced without warning, affects thousands of workers. The changes, long overdue according to critics, will take effect immediately.

Before returning the final rewrite, scan it for `—` and `–`. Any hit means the draft isn't done. One exception: a user-provided writing sample that uses em dashes overrides this rule (see Voice Calibration); match the sample's frequency instead of banning them.

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The document explicitly discloses automatic updates, so this is not entirely hidden behavior, but ordinary publishing triggers replacement of installed skill code without per-update approval. This creates a supply-chain trust boundary: a compromised upstream release could run with access to publishing credentials and user files.

Content

No source excerpt is available for this finding.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 26)May include surrounding context.

md
| **check-update.js** | 同上 | 推送前核对 / 自升级(由 `push.js` 调用) |

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 142)May include surrounding context.

md
`push.js` 推送前对照远端最新版;落后则升级技能文件并继续推送(**保留 `config.json`**)。同目录 24 小时检查一次。网络失败不阻断推送。

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 180)May include surrounding context.

md
在 **aigc-web-push 目录**执行 `push.js`;通道为 `wechat` 或 `cloud`。下方路径示例假设作品目录相对技能目录为 `../works/20261117-demo/`(按实际路径调整)。

The skill implements default-enabled self-updating that overwrites installed skill files, beyond its stated content-generation and publishing purpose.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

The updater trusts the registry's latest release and writes downloaded files directly over installed code without an independently authenticated release signature or user approval. HTTPS alone does not protect against upstream compromise, and unrestricted redirects permit transport downgrade; sequential non-atomic writes can also leave partially updated code after failure.

Content

No source excerpt is available for this finding.

Unconditional activation applies the rewriting skill outside explicit humanization requests.

Medium
Category
Trigger Abuse
Confidence
95% confidence
Finding

L003 sets alwaysApply: true, making the skill active regardless of whether the user requested text humanization. Although the description identifies an editing purpose, it does not constrain this unconditional activation, so the rewriting rules may affect unrelated tasks.

Content

No source excerpt is available for this finding.

The unconditional `alwaysApply: true` setting applies the editing skill outside requests to humanize text.

Medium
Category
Trigger Abuse
Confidence
93% confidence
Finding

L002 describes a text-editing use case, but L003 enables the skill unconditionally rather than limiting activation to that use case. This can introduce its rewriting and style constraints into unrelated tasks without a request to humanize content.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx remotion'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
93% confidence
Finding

Although this is not an MCP server, npx skills can download and execute an unpinned npm package, and the referenced remote skill repository is also unpinned. A compromised package or changed skill revision could introduce malicious code or agent instructions into a workflow with workspace and publishing access.

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 243)May include surrounding context.

md
- 公众号 / 云电脑未指定时用 `-` → 配置里 `selected: true` 项;也可传名称或 id。
- 云电脑指定环境/账号时按名称解析,歧义则失败并打印候选。
- API 基址:`apiBase`(缺省 `https://api.pcloud.ac.cn/openAccessService`)。
- **云电脑入队即完成**:勿轮询、勿因「还没发出」重推。
- **公众号超时**:可视为已受理,勿狂推;请用户看服务通知或草稿箱。

L249–L259 provide a bulk draft-deletion capability unrelated to the manifest's stated content-generation and publishing purpose.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The skill documents an authenticated cleanupDrafts request to empty a WeChat account's draft box. Unlike generating or pushing content, this is a destructive account-maintenance operation, and the request contains no draft identifiers to restrict deletion to content created by the current task.

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 251)May include surrounding context.

清空草稿箱

POST https://api.pcloud.ac.cn/openAccessService:

json
{

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 · config.example.json (reported line 4)May include surrounding context.

json
{
  "_fileRole": "字段说明 + 示例。正式 config.json 只保留向导输出的业务键(去掉本文件中的 _fileRole / fieldsHelp)。",
  "fieldsHelp": {
    "apiBase": "缺省 https://api.pcloud.ac.cn/openAccessService",
    "openId": "扫码得到的微信 openId;禁止编造或共用他人 config",
    "skillKey": "向导签发,与 openId 绑定;推送与音乐搜索必带;重新扫码会轮换",
    "autoUpdate": "可选;false 关闭推送前自动升级(默认开启)",

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 · config.example.json (reported line 14)May include surrounding context.

json
},
  "version": 1,
  "description": "内容生成与全网推送(公众号 + 云电脑)",
  "apiBase": "https://api.pcloud.ac.cn/openAccessService",
  "openId": "oXXXXXXXXXXXXXXXXXXXXXXXXXXX",
  "skillKey": "请用向导扫码后生成,勿手写",
  "autoUpdate": true,

The HTML-generation rules are configured to apply unconditionally, without an HTML-task activation boundary.

Medium
Category
Trigger Abuse
Confidence
91% confidence
Finding

Although the description identifies an HTML generator, alwaysApply: true activates these instructions without restricting them to HTML-generation tasks. This can impose the file's strict output and formatting requirements on unrelated requests.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% 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 · push-to-cloud.js (reported line 28)May include surrounding context.

js
const path = require('path');
const { URL } = require('url');

const DEFAULT_API = 'https://api.pcloud.ac.cn/openAccessService';
const DIR = __dirname;
const TARGET_SCOPES = new Set(['all', 'environment', 'account']);
const PUBLISH_MODES = new Set(['draft', 'publish', 'default']);

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% 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 · push-to-wechat-mp.js (reported line 17)May include surrounding context.

js
const path = require('path');
const { URL } = require('url');

const DEFAULT_API = 'https://api.pcloud.ac.cn/openAccessService';
const DIR = __dirname;

function readJson(name) {

The music-search workflow sends project-derived search terms and an account identifier to an external service without a user-facing privacy warning.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The skill instructs the agent to search api.pcloud.ac.cn using music keywords together with openId and skillKey from config.json. Although it documents the endpoint and authentication headers, it does not instruct the agent to notify the user that topic-derived terms and their account identifier will be transmitted externally; existing configuration alone does not disclose that privacy impact.

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 · remotion-video.md (reported line 206)May include surrounding context.

md
**音乐搜索**

`GET https://api.pcloud.ac.cn/proxyService/heygen/sounds?query=%E5%85%B3%E9%94%AE%E8%AF%8D&limit=10%60

请求头(与推送相同,取自本技能目录 `config.json`):

MCP server referenced without pinned version: 'npx remotion'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
91% confidence
Finding

Although this is not an MCP server, npx skills add remotion-dev/skills can download and execute an unpinned registry package, then install skill instructions from an unpinned repository revision. A compromised package release or upstream repository could therefore execute code or introduce malicious instructions with the agent's available permissions; requiring installation as part of the workflow increases exposure.

Content

No source excerpt is available for this finding.