Back to skill

Security audit

12321 骚扰电话举报

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly matches its stated goal of automating a harassment-phone report, but it handles real reports, SMS verification, personal phone data, and optional consent in ways that need careful review before use.

Install only if you are comfortable with an agent automating a real government-style complaint workflow. Before using it, remove the hard-coded phone number, require explicit confirmation before SMS and final submission, leave optional service consent unchecked unless the user opts in, avoid bypassing site safeguards, and ensure temporary CAPTCHA files are securely deleted.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • 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
Findings (5)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:104
Finding

<![CDATA[Unescaped Values Are Interpolated into Browser-Executed JavaScript]]>

Content
View full analysis
{ // 1. Fill in CAPTCHA const el=document.getElementById('w_code'); el.removeAttribute('onkeyup'); el.removeAttribute('onpaste'); el.removeAttribute('oncontextmenu'); el.value='{captcha}'; el.dispatchEvent(new Event('input',{bubbles:true})); // 2. Validate value const w=el.value; if(!w){return JSON.stringify({error:'CAPTCHA is empty; refresh the page and retry'});} // 3. Fill in phone number const s=Object.getOwnPropertyDescriptor(window.HTMLInputElement.prototype,'value').set; const el2=document.getElementById('phone'); s.call(el2,'{phone}'); el2.dispatchEvent(new Event('input',{bubbles:true})); el2.blur(); // 4. Trigger SMS document.getElementById('hq2').click(); return JSON.stringify({ w:document.getElementById('w_code').value, p:document.getElementById('phone').value, t:document.getElementById('Time')?.textContent?.trim() }); } ``` Additional vulnerable substitutions include: ```javascript s.call(pc,'{sms_code}'); s.call(sp,'{report_phone}'); ``` ```javascript document.getElementById('d241').value='{YYYY-MM-DD}'; document.getElementById('d242').value='{HH:mm}'; document.getElementById('d243').value='{long_time}'; ``` ```javascript ts.call(sc,'{来电描述}'); ``` ### Technical Analysis The Skill constructs browser-evaluation scripts by inserting CAPTCHA, phone, SMS code, reported number, date, time, enum, and free-text description values directly into single-quoted JavaScript string literals. It does not escape quotation marks, backslashes, line terminators, or other JavaScript syntax. Several values originate directly from user messages. In particular, the free-text call description can naturally contain arbitrary punctuation. If an attacker-controlled value contains a quote followed by JavaScript sy ...[truncated 1380 chars]
Remediation
View remediation
{ const captchaElement = document.getElementById('w_code'); captchaElement.value = captcha; captchaElement.dispatchEvent(new Event('input', { bubbles: true })); const setter = Object.getOwnPropertyDescriptor(HTMLInputElement.prototype, 'value').set; const phoneElement = document.getElementById('phone'); setter.call(phoneElement, phone); phoneElement.dispatchEvent(new Event('input', { bubbles: true })); } ``` - If structured arguments are unavailable, encode every substituted value with `JSON.stringify` and insert the resulting JSON literal without surrounding it with manually written quotes. - Apply strict allowlists before evaluation: - CAPTCHA and SMS code: expected character set and exact length. - Phone numbers: only formats accepted by the service. - Date and time: validated calendar and time formats. - Enum fields: exact predefined values. - Description: bounded length and normalized text. - Add tests containing quotes, backslashes, newlines, Unicode separators, and JavaScript-like payloads to verify that values remain data rather than executable source. ]]>

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:61
Finding

<![CDATA[Hard-Coded Personal Phone Number Is Used as the SMS Verification Default]]>

Content
View full analysis
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
SKILL.md:182
Finding

<![CDATA[Optional Protection-Service Consent Is Enabled Without Explicit Authorization]]>

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:17
Finding

<![CDATA[Website Countdown and Input Safeguards Are Deliberately Disabled]]>

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Note
Location
SKILL.md:39
Finding

<![CDATA[CAPTCHA Data Is Written to Predictable Files Without Cleanup]]>

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (8)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The README advertises automatic form submission and SMS-code triggering without an explicit warning that the skill may send data to a third-party government reporting site, initiate verification messages, and potentially submit a complaint on the user’s behalf. In this skill’s context, that omission increases risk because the workflow handles phone numbers, complaint details, and verification steps tied to real-world reporting actions.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The trigger phrases include generic requests such as “填写举报表单” and “打开12321”, which can plausibly appear in ordinary conversation and unintentionally invoke an automation that drives a browser and progresses a complaint workflow. In this skill’s context, accidental activation is more dangerous because the automation can bypass countdowns, initiate SMS verification, and prepare or submit a real external report involving personal data.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The trigger phrase '打开12321' is broad enough to invoke the skill in contexts where the user may only want information or navigation, not automated complaint submission. Because this skill handles sensitive personal data and initiates a multi-step reporting workflow, accidental invocation can lead to unintended collection or action.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill is explicitly designed to collect, process, and submit sensitive user data including phone numbers, captcha values, and SMS verification codes, but it provides no privacy warning, consent checkpoint, or data-handling minimization guidance. Because these artifacts are authentication-related and personally identifying, silent handling raises risk of accidental disclosure, misuse, or overcollection.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill directs the agent to use a local Node.js command to transform captcha data and write files to disk, which unnecessarily expands the agent's capability from browser automation into local code execution and filesystem access. In a hostile or modified workflow, this pattern can be repurposed to run arbitrary commands or persist sensitive artifacts locally, increasing attack surface beyond what is needed for form submission.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill prepopulates and echoes sensitive phone data in confirmation messages, including a hardcoded default number and report details, which creates unnecessary disclosure risk and may cause submission with the wrong identity data. In context, this is especially concerning because the workflow also handles SMS verification and complaint filing, so misuse could expose personal data or file reports using stale information.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

The manifest states the skill automates reporting on the 12321 site, including captcha recognition, but the documented fallback performs OCR directly against an image URL. While related to captcha solving, this introduces an additional remote-content processing mode not clearly justified or scoped in the manifest description.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

Sending the captcha image and recognized text back through chat exposes verification artifacts in a secondary channel, which can increase leakage risk if chat history is retained, shared, or visible to others. Although captcha content is short-lived, treating verification material as normal chat content without warning or minimization is unsafe.

Content

No source excerpt is available for this finding.