Back to skill

Security audit

web-session-miner

Security checks for vulnerabilities and agentic risk

Overview

The skill is an openly described logged-in browser scraping workflow, but it needs review because it reuses live sessions broadly and markets avoiding official API billing.

Install only if you are comfortable letting the agent use a logged-in browser profile to access and copy data your account can see. Use a dedicated browser profile and least-privileged account, confirm the exact target site and account before each run, avoid bulk scraping unless the site permits it, and prefer official APIs when the provider requires them for programmatic access.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (12)

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The README’s compliance language conflicts with earlier marketing that explicitly touts using a logged-in web session to avoid official API per-call billing. That contradiction normalizes using browser-session reuse to obtain data through channels the provider may have intended to meter separately, increasing the likelihood of ToS violations, account sanctions, and abusive scraping workflows. In this skill’s context, the danger is heightened because the core purpose is automated extraction from authenticated web sessions rather than incidental browser automation.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill explicitly instructs reuse of an already logged-in browser session and cookies to programmatically extract data from authenticated websites. This is dangerous because it operationalizes credential/session piggybacking: the agent can act with the user's account privileges and access subscription-gated or sensitive data without granular re-authentication or per-request consent.

Content

No source excerpt is available for this finding.

Ssd 3

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The instructions tell the agent to inherit browser cookies and then use snapshot/full-text extraction on authenticated pages, which can capture far more information than the intended target fields. Full-page text collection from logged-in pages risks overcollection of personal data, account data, hidden identifiers, and unrelated sensitive content exposed in the user's session.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The README encourages reusing a live logged-in browser session but does not prominently warn that this grants the skill access to whatever the current account can view, potentially exposing cookies, personal data, enterprise data, and paid-account entitlements. In this context, that omission is especially dangerous because the skill’s value proposition depends on authenticated session reuse across third-party sites, making session mix-ups, over-collection, and accidental disclosure more likely.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

Broad trigger phrases like '查企业', '查老板', or '网页抓取' create ambiguous activation boundaries and can cause the agent to invoke a high-risk skill in situations where the user did not specifically consent to authenticated browser-session reuse. Because this skill can act through a live logged-in browser and collect sensitive commercial data, overbroad invocation materially increases the chance of unintended scraping, privacy violations, or actions taken under the wrong account context.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrases are broad business terms like company lookup, shareholder lookup, due diligence, and web scraping, which can cause the skill to activate for routine user queries without clear intent to reuse authenticated browser sessions. Because this skill is high-risk and session-sensitive, overbroad invocation materially increases the chance of accidental data access or scraping actions.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The document discusses compliance boundaries but fails to clearly warn that reusing browser login state exposes sensitive account context, including cookies, session-bound data, and potentially broad authenticated content. Users may understand legal/commercial limits yet still miss the privacy and account-security implications of granting an automation agent access to their live session.

Content

No source excerpt is available for this finding.

Ssd 2

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The repeated framing of the workflow as a way to avoid API points or costs encourages deliberate circumvention of the provider's intended metering and access model. Even if the same data is visible in the browser, packaging the skill around defeating paid programmatic interfaces raises abuse risk and can expose users and operators to contractual, legal, and account-enforcement consequences.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill presents itself as compliance-limited, but elsewhere explicitly markets the workflow as a way to avoid official API/MCP billing by scraping authenticated web pages. That contradiction can normalize misuse, encourage terms-of-service circumvention, and lead the agent to help automate access patterns the site operator intended to meter separately.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The entire template is written in Chinese and instructs authors to use it as the standard site-adapter template, but it does not offer any language or locale choice. Under the policy, a skill artifact that imposes a specific language without user opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
79% confidence
Finding

Line L47 prescribes a specific date format and Chinese unit labels such as '万美元/万元' as a general rule, and the overall skill is written as a fixed Chinese-localized workflow with no opt-in or alternative locale guidance. This can violate language/locale policy when the skill is used in broader contexts and does not explicitly limit itself to a China-specific compliance or locale-only use case.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

The entire skill file is written in Chinese and provides mandatory procedural instructions such as submission checks, but it does not indicate that another language is acceptable or that Chinese is optional. Under the natural-language policy rule, a forced language/locale without user opt-in can be a policy concern.

Content

No source excerpt is available for this finding.