ClawHub
Back to skill

Security audit

binancesquare

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly performs the advertised feed fetch, but it needs Review because it sends requests to a non-obvious bmwweb.cc endpoint with a fixed cookie and UUID while presenting the data as Binance Square content.

Install only if you are comfortable with the skill contacting www.bmwweb.cc, sending the embedded bnc-uid/Bnc-Uuid values, and trusting that endpoint for financial social-feed content. Prefer a version that uses a documented official Binance endpoint or clearly discloses the third-party source, removes fixed identifiers, and sanitizes remote text before printing.

A.I.G

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 (2)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/binance_square.py:9
Finding

Unverified Third-Party API Use with a Hardcoded Tracking Identifier

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Note
Location
scripts/binance_square.py:67
Finding

Terminal Control-Sequence Injection Through Untrusted Feed Content

Content
View full analysis
7} | {comments:>5}") ``` ### Technical Analysis The `content` and `authorName` fields originate from an external API response and are printed directly to the terminal. Truncating these strings by character count does not remove ANSI escape sequences, C0/C1 control characters, carriage returns, backspaces, bidirectional text controls, or terminal hyperlink sequences. If an attacker can publish a specially crafted feed entry, or if the external API is compromised, terminal-emulator control sequences can be placed in either field. When the script prints the values, a compatible terminal may interpret the sequences instead of displaying them as ordinary text. This code does not pass the values to a shell and therefore does not directly provide shell-command execution. The confirmed risk is manipulation of terminal output and terminal state. ### Attack Path 1. An attacker creates a feed post or author name containing ANSI escape or other terminal control sequences. 2. Alternatively, a compromised API injects those sequences into an otherwise legitimate response. 3. The malicious entry passes the recency and retail-user filters. 4. The script copies the remote `content` and `authorName` values without sanitization. 5. The `print` call writes the values to the user's terminal. 6. The terminal interprets the sequences, allowing the attacker to hide text, overwrite displayed rows, change colors, produce deceptive hyperlinks, or otherwise falsify the apparent output. ### Impact Assessment No ...[truncated 525 chars]
Remediation
View remediation

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (7)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
96% confidence
Finding

The skill clearly instructs the agent to perform an external network request but does not declare any tool scope, permissions, or allowed-tools constraints. This creates an over-broad execution surface where a caller or host may not realize the skill transmits data externally, increasing the risk of unintended outbound requests and weak reviewability.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill includes concrete request headers and cookie values for an external API but does not warn that invoking it transmits request metadata and cookie-like identifiers off-platform. This is dangerous because users or operators may unknowingly send persistent identifiers or session-related values to a third party, creating privacy, compliance, and possible credential-handling risks.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The usage guidance says to directly invoke the skill, without defining when it should or should not run, what user consent is needed, or what prerequisites apply. That makes accidental or overly broad invocation more likely, especially since the skill performs external requests and handles request metadata including cookies.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

This code hard-codes a Chinese locale in natural-language and protocol fields, including the Referer path '/zh-CN' and cookie 'lang=zh-CN'. The file also presents all user-facing text in Chinese, with no opt-in or documented region-specific requirement, which matches the language/locale policy violation criteria.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
91% confidence
Finding

The script sends outbound requests to a third-party domain using hardcoded headers, including a static cookie and UUID, without validating provenance or making this data flow explicit to the user. This creates privacy and trust risks because using a non-official-looking host to retrieve Binance-related content could expose request metadata and encourage transmission to an unverified external service.

Content

Scanner excerpt · scripts/binance_square.py (reported line 29)May include surrounding context.

python
url = "https://www.bmwweb.cc/bapi/composite/v9/friendly/pgc/feed/feed-recommend/list"
    payload = {"pageIndex": page, "pageSize": size, "scene": "web-homepage", "contentIds": []}
    try:
        r = requests.post(url, headers=HEADERS, json=payload, timeout=10)
        r.raise_for_status()
        d = r.json()
        return d.get("data", {}).get("vos", []) if d.get("success") else []

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The documented request headers hard-code a Chinese locale via the Referer path '/zh-CN/square' and the cookie value 'lang=zh-CN'. This is a natural-language locale constraint presented without user opt-in or explanation, which can violate language or locale policy expectations.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The manifest description limits the output to 内容、浏览量、评论数到表格, while the actual printed table contains additional fields 作者 and 时间. This is a mild description-behavior mismatch because the code exposes more data than the stated output scope, even though it is related to the same posts.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.