Back to skill

Security audit

双币赢分析

Security checks for vulnerabilities and agentic risk

Overview

This skill is a review concern because it gives actionable crypto options recommendations from an unauthenticated plaintext remote service that is not clearly disclosed.

Review this carefully before installing. It does not appear to steal files or install persistence, but its trading recommendations depend on a remote plaintext HTTP service at a raw IP address. Do not treat its output as verified investment advice, and prefer a version that uses HTTPS, documents the provider, validates response integrity, shows data timestamps, and requires explicit user confirmation before producing actionable recommendations.

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
run.py:10
Finding

Unauthenticated Plaintext API Controls Financial Recommendations

Content
View full analysis

Vulnerability Details

File Location: run.py, lines 10-23
Vulnerability Type: Untrusted remote response over plaintext HTTP
Risk Level: High

Vulnerable Code:

python
API_URL = "http://43.156.132.183:30080/api/analysis"

def get_data():
    t = int(time.time())
    resp = requests.get(f"{API_URL}?t={t}", timeout=30)
    return resp.json()

def main():
    data = get_data()
    
    if data.get('success'):
        print(data.get('output', ''))
    else:
        print("Error:", data.get('error', 'Unknown error'))

Technical Analysis

The skill retrieves its financial analysis from a raw IP address over unencrypted HTTP. The request has no transport confidentiality or integrity, endpoint authentication, application-level response signature, or other mechanism capable of proving that the response originated from an authorized analysis provider.

The implementation also accepts the response as JSON and directly prints the remotely controlled output value without checking the HTTP status, content type, response size, expected schema, or provenance. Consequently, an on-path attacker or a party controlling the remote server can replace or modify the displayed recommendation.

Although SKILL.md describes local decision logic involving RSI, skew, DVOL, IVP, and IVK, none of these calculations are implemented in the reviewed script. The effective analysis and recommendation logic is therefore controlled entirely by the remote service and can change after the skill package has been audited.

This is not remote code execution in the reviewed implementation: the response is printed rather than passed to exec, a shell, or another local execution primitive. The direct security impact is manipulation of output integrity.

Attack Path

  1. A user invokes one of the triggers configured for the DualWin skill.
  2. The skill executes run.py.
  3. run.py sends an HTTP request to `43.156.132. ...[truncated 1132 chars]
Remediation
View remediation

Remediation Suggestions

  1. Replace the plaintext endpoint with an HTTPS endpoint using a stable, controlled domain and valid certificate verification.
  2. Authenticate requests and require cryptographically signed responses. Verify signatures before using any returned analysis.
  3. Call resp.raise_for_status() and reject unexpected HTTP status codes.
  4. Require the expected JSON content type and enforce a strict response schema, including type, length, and allowed-value checks for every field.
  5. Apply a conservative maximum response size before parsing or displaying remote content.
  6. Treat all remote text as untrusted data. Ensure downstream agents or automation cannot interpret returned text as executable instructions.
  7. Implement the documented indicator calculations locally where feasible, using independently authenticated market-data sources, so the recommendation can be reproduced and audited.
  8. If server-side analysis remains necessary, clearly disclose the remote provider and verify returned market values and recommendations against independent data.
  9. Add explicit failure handling for connection errors, malformed JSON, schema violations, and authenticity-verification failures; do not display unverified recommendations.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • 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 (7)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger scope is broad enough to activate on generic BTC/ETH trading or investment questions, which can cause the skill to engage outside a narrowly intended context. In a financial skill that presents specific trading recommendations and 'win rate' outputs, unintended activation increases the chance users receive automated, high-risk guidance without explicitly requesting this specialized analysis.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill markets itself as a professional decision assistant using real-time indicators and automatically identifying optimal subscription plans, but it does not clearly warn that stale, missing, or inaccurate market data could materially distort the recommendations. In the context of leveraged or options-like products, users may over-trust the automation and act on outputs that appear authoritative despite data dependency and model uncertainty.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill makes an outbound request to a hard-coded external IP address over plain HTTP and provides no user-facing disclosure that data or execution context may be sent off-device. This is risky because traffic can be intercepted or modified in transit, and users are not informed that the skill depends on or communicates with a remote service.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrase "双币赢" is very broad and can match ordinary user discussion about a financial product rather than an explicit request to invoke this skill. Because this skill executes a Python program that provides trading analysis, accidental activation could cause unintended market-data processing or unsolicited trading recommendations in a sensitive finance context.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The English triggers "dualwin", "btc dual", and "eth dual" are ambiguous and short, making them prone to matching casual conversation or unrelated finance queries. In this skill, ambiguous activation is more concerning because the action launches an executable Python analysis workflow, so accidental invocation may produce unrequested financial guidance or consume privileged runtime resources.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

The skill content, invocation phrase, and output template are presented entirely in Chinese, with no indication that users may choose another language. This can violate language/locale policy when a skill imposes a language without opt-in or documented regional justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The inline natural-language comment is written in Chinese, while the file otherwise does not indicate that the skill is region-specific or that users can opt into a language/locale. This can indicate an implicit language policy choice without documented user choice or justification.

Content

No source excerpt is available for this finding.