Back to skill

Security audit

Wyckoff Screen

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly matches a stock-screening purpose, but it hides a hardcoded market-data token and custom plaintext HTTP endpoint that can affect financial recommendations.

Install only after reviewing the data-provider code. Expect the skill to contact external stock-data services, write local SQLite caches and dated result files, and produce Chinese stock-screening output. The hardcoded API token and undocumented plaintext HTTP endpoint should be removed or replaced with a user-provided secret and trusted HTTPS provider before use.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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 (2)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/ts_data.py:10
Finding

Hardcoded API Token Transmitted to a Custom Plaintext HTTP Endpoint

Content
View full analysis
pd.DataFrame | None: # 方法1: tushare try: df = get_quote_history_ts(code, start_date='20200101', end_date='20500101') if df is not None and len(df) >= 60: return df except Exception: pass ``` The primary scanner calls that fetcher for each stock: ```python try: df = fetch_stock_data(code) time.sleep(REQUEST_DELAY) if df is None or len(df) < 60: continue except: time.sleep(REQUEST_DELAY) continue ``` ### Technical Analysis The source code embeds a reusable API token and assigns it to a Tushare client. It then overrides private client fields so that requests are sent to `http://140.143.209.128:5000`. Because the endpoint uses plaintext HTTP rather than authenticated HTTPS, the connection does not provide transport confidentiality, server authentication, or integrity protection. A network-positioned attacker may be able to observe the token and query parameters, replay the credential, redirect traffic, or modify returned market data. Use of a bare IP address and private Tushare client attributes also bypasses the normal provider configuration and makes the endpoint harder to authenticate and audit. The path is active: `screen.py` calls `fetch_stock_data()`, which attempts this Tushare ...[truncated 1616 chars]
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
SKILL.md:99
Finding

Unpinned and Incompletely Documented Third-Party Dependencies

Content
View full analysis
= 60: return df except Exception: pass ``` ### Technical Analysis The documentation recommends installing dependencies without exact versions or integrity hashes. Package resolution can therefore change over time, producing non-reproducible installations and exposing users to future compromised, malicious, or unexpectedly incompatible releases. The dependency inventory is also incomplete. The implementation directly relies on packages including `tushare`, `baostock`, `pandas`, and `numpy`, but they are not all declared in the documented dependency section. In particular, importing `screen_data.py` imports `ts_data.py`, which requires Tushare even before fallback behavior is evaluated. Python packages execute co ...[truncated 2094 chars]
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
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (22)

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

This finding alleges undisclosed external HTTP usage and a hardcoded API token; if present, that would be a significant security issue because it introduces secret handling and external data exfiltration not declared in the skill. However, the provided SKILL.md content itself does not show such a token or endpoint, so the concern is plausible only if detected from related code rather than this file alone.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

This finding alleges undisclosed external HTTP usage and a hardcoded API token; if present, that would be a significant security issue because it introduces secret handling and external data exfiltration not declared in the skill. However, the provided SKILL.md content itself does not show such a token or endpoint, so the concern is plausible only if detected from related code rather than this file alone.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

This finding alleges undisclosed external HTTP usage and a hardcoded API token; if present, that would be a significant security issue because it introduces secret handling and external data exfiltration not declared in the skill. However, the provided SKILL.md content itself does not show such a token or endpoint, so the concern is plausible only if detected from related code rather than this file alone.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
82% confidence
Finding

The skill describes local SQLite persistence and references scripts that update and store market data, but it does not declare any explicit tool scope or permissions. Undeclared file-write capability increases risk because the runtime may allow disk modification beyond what users expect, reducing transparency and auditability.

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 natural-language investment prompts such as '今日买什么' and '帮我看看有什么可以买的', which can overlap with ordinary conversation. Over-broad invocation is risky because it may cause the skill to run unexpectedly, perform undeclared market scans or local writes, and present investment recommendations when the user did not intend to invoke this workflow.

Content

No source excerpt is available for this finding.

Ae2

Medium
Category
analysis-evasion
Confidence
90% confidence
Finding

Artifact content does not match its filename extension

Content

No source excerpt is available for this finding.

Ae2

Medium
Category
analysis-evasion
Confidence
90% confidence
Finding

Artifact content does not match its filename extension

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file content is a Chinese-language user-facing message and provides no indication that the user can choose another language or opt in to this locale. This can violate a language/locale policy when a skill imposes a specific language by default.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The module documentation says it scans '全市场' to find candidates, which implies the full A-share market. However, get_all_codes() restricts the universe to codes starting with 000, 001, or 002, and logs that only '深交所主板+中小企业板' are included, so the documented scope contradicts the implemented behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This code file contains natural-language strings that force a specific language/locale for usage instructions and output. Under the policy, language constraints should not be imposed without giving the user a choice or clearly documenting a justified region-specific requirement.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

The only natural-language descriptions in the file are written in Chinese, with no indication that the user can choose another language or locale. This may violate the language/locale policy when a skill effectively constrains communication to a specific language without explicit opt-in or documented regional justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The file's title and all user-facing strings are hardcoded in Chinese, indicating the skill enforces a specific language/locale. The policy allows this only when the skill offers user opt-in or clearly documents that it is intentionally region- or locale-specific; this file does not provide such a choice or justification in user-facing behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The script creates and persists a local SQLite database containing downloaded market data and screening results, which exceeds the manifest’s described behavior of returning ranked stock candidates. Undisclosed persistent storage increases data-retention and side-effect risk, especially in agent environments where users may expect a read/compute-only action.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The file hardcodes a live Tushare API token and rewires the client to use a custom external endpoint, which creates two concrete risks: secret exposure and redirection of market-data requests to an untrusted service. Because the endpoint uses plain HTTP, requests and responses can be intercepted or modified in transit, and anyone with access to the code can reuse the embedded token.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
99% confidence
Finding

Embedding the API token directly in code exposes credentials to every reader of the repository, logs, backups, and downstream package consumers. This enables unauthorized use of the account, quota exhaustion, billing abuse, or access revocation, and the token is immediately actionable because it is used to initialize network access.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The adapter overrides the API client's base URL to an explicit outbound HTTP server, causing stock-query traffic to be sent to a non-default remote host without transport security. In a stock-screening skill, this is especially risky because manipulated or intercepted market data can directly influence buy/sell recommendations, creating integrity as well as confidentiality concerns.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This Python file contains user-facing natural-language strings and documentation entirely in Chinese, including the module header and verdict/warning messages returned by scoring functions. Under the policy, forcing a specific language without user opt-in or a documented region-specific justification is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The only natural-language content in the file is Chinese text, which indicates a fixed language/locale presentation. There is no surrounding indication of user opt-in or a documented reason that this skill is intentionally locale-specific, so it may violate the language-choice policy.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The manifest describes a stock-screening skill that outputs ranked candidates and workflow explanation, but this file also initializes and writes a SQLite database and saves daily text reports to disk. Persistent local storage is additional behavior not reflected in the stated description, even though screening itself is expected.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

This code performs a remote market-data fetch via efinance, which is a network operation, but there is no confirmation prompt, user-facing print/log message, or warning at the point of use. Although comments describe fallback behavior, they do not disclose to the user that external network access will occur.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

The script writes screening output to a dated text file on disk even though the stated skill behavior is to produce a ranked candidate list. This hidden file-write side effect can surprise operators, accumulate historical outputs, and expose potentially sensitive usage or strategy information to other local processes or users.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

The natural-language strings and instructions are presented only in Chinese, including a mandatory-sounding note, with no indication that users may choose another language. This can violate language/locale policy when a specific language is effectively forced without opt-in or justification.

Content

No source excerpt is available for this finding.