Back to skill

Security audit

Wyckoff Trading Agent

Security checks for vulnerabilities and agentic risk

Overview

This appears to be a real Wyckoff/A-share trading skill, but its setup and credential handling are risky enough that users should review it carefully before installing.

Review this before installing. Prefer a pinned, verified package in a virtual environment; do not run the curl-to-bash installer; avoid placing passwords or API keys directly in shell commands or MCP config files; and understand that the skill can register a persistent MCP tool, read saved Wyckoff credentials, and guide portfolio-changing commands. The TickFlow paid referral link should also be treated as commercial steering unless clearly disclosed by the publisher.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
Findings (4)

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:25
Finding

Mutable Remote Installation Script Is Executed Directly by a Shell

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Error
Location
SKILL.md:68
Finding

Unpinned Third-Party Package Is Installed and Registered as an MCP Tool

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
rules/cli-setup-guide.md:17
Finding

Passwords and API Keys Are Passed Through Command-Line Arguments and Persisted Locally

Content
View full analysis
` (credentials will be persisted, no need to login again). 3. Verify with `wyckoff auth status`. ``` `SKILL.md:40-54`: ```markdown - **tushare_token** missing → guide: - Tushare token is free, get it from https://tushare.pro/ after registration. - Then run: `wyckoff config tushare ` - **tickflow_api_key** present → OK. - **tickflow_api_key** missing → guide: - TickFlow provides real-time and historical A-share data. - Purchase at: https://tickflow.org/auth/register?ref=5N4NKTCPL4 - Then run: `wyckoff config tickflow ` ### 0.4 Model Check Run `wyckoff model list`. - **Has models** → proceed to Step 1. - **No models** → guide: `wyckoff model add` (interactive) or `wyckoff model set --model `. ``` `rules/cli-setup-guide.md:17-27`: ```markdown 1. Open https://wyckoff-analysis.pages.dev/ 2. Register with email and password. 3. Login via CLI: ```bash wyckoff auth login ``` 4. Verify: `wyckoff auth status` Note: Credentials are automatically saved to `~/.wyckoff/wyckoff.json`. When the login token expires, the CLI will re-login automatically — no need to enter credentials again. ``` ### Technical Analysis The documented commands place passwords, market-data tokens, and model-provider API keys directly in process arguments. Depending on the operating system and shell, these values may be retained in shell history, captured in terminal logs, exposed in process listings, or collected by monitoring and diagnostic tools. The guide additionally stat ...[truncated 1509 chars]
Remediation
View remediation

other

Note
Location
SKILL.md:43
Finding

Setup Flow Steers Users to a Paid Service Through an Undisclosed Referral Link

Content
View full analysis
` Note: At least one data source (tushare or tickflow) must be configured. Both are recommended for better coverage. ``` `rules/cli-setup-guide.md:37-43`: ```markdown ### TickFlow (paid, real-time + historical) 1. Purchase at https://tickflow.org/auth/register?ref=5N4NKTCPL4 2. Copy your API key from the dashboard. 3. Configure: ```bash wyckoff config tickflow ``` ``` ### Technical Analysis The project repeatedly directs users to a paid provider through a URL containing the referral identifier `ref=5N4NKTCPL4`. No commercial, affiliate, or referral relationship is disclosed. The documentation states that only one data source is required and that Tushare is an available alternative. Therefore, steering users to a referral-based paid registration is not technically necessary for the Skill's declared analysis functionality. The issue is one of transparency and potentially biased recommendation rather than direct code execution. ### Attack Path 1. A user follows the guided setup process. 2. The Skill presents the referral-bearing TickFlow registration URL as part of data-source configuration. 3. The user visits the tracked URL and may purchase the service. 4. The referral identifier attributes the registration or purchase to the link owner without the Skill disclosing that tracking or possible benefit. ### Impact Assessment The user may be influenced to purchase an optional service without being informed of the referral relationship. The provider receives ...[truncated 167 chars]
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
Findings (26)

Chaining Abuse

High
Category
Tool Misuse
Confidence
99% confidence
Finding

The | bash chaining pattern turns remote content into immediate shell execution, removing opportunities for inspection and magnifying the consequences of source compromise. In a skill that actively guides installation and setup, this is more dangerous than incidental documentation because users are likely to follow the exact command, leading to arbitrary command execution with the user's privileges.

Content

Scanner excerpt · SKILL.md (reported line 25)May include surrounding context.

text
  Or one-line install:

curl -fsSL https://raw.githubusercontent.com/YoungCan-Wang/Wyckoff-Analysis/main/install.sh | bash

text
After install, verify with `wyckoff --version` again.

Chaining Abuse

High
Category
Tool Misuse
Confidence
98% confidence
Finding

The use of a pipe into bash is a chaining pattern that turns downloaded content directly into shell commands with no inspection boundary. This materially increases exploitability because any malicious modification to the downloaded script becomes immediate command execution under the user's privileges.

Content

Scanner excerpt · rules/cli-setup-guide.md (reported line 12)May include surrounding context.

Alternative (one-line):

bash
curl -fsSL https://raw.githubusercontent.com/YoungCan-Wang/Wyckoff-Analysis/main/install.sh | bash

Verify: wyckoff --version

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The guide tells users to pass email and password directly as CLI arguments, which can expose secrets through shell history, process listings, terminal logging, and audit systems. In shared or monitored environments, this can lead to credential theft and unauthorized account access.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill recommends a one-line curl ... | bash installation flow that downloads and immediately executes a remote script without verification. This creates a supply-chain execution risk: if the remote content, transport path, repository, or account is compromised, users may run arbitrary code on their machine. Because this is positioned as a primary install path in a setup guide, the skill materially increases the likelihood of unsafe execution.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill instructs users to pass email and password directly on the command line via wyckoff auth login <email> <password>. Command-line arguments are commonly exposed through shell history, process listings, terminal logging, and remote session audit trails, which can leak credentials to other local users or monitoring systems. In this setup-oriented skill, that risk is real because the instruction is explicit and normalized as the recommended login path.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

L145 mandates printing 当前北京时间:YYYY-MM-DD HH:MM(UTC+8), and the fixed output contract at L186-L192 continues this Chinese-only format. The file does not state that the user can choose language or that the locale restriction is required for a region-specific compliance reason, so this is a natural-language locale policy issue.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The default prompt authorizes a very broad chain of actions: detecting local environment, guiding installation/configuration, fetching data from multiple sources, processing CSV/images/portfolio input, and directly executing CLI operations. Without explicit trigger boundaries or user-confirmation gates for sensitive steps, the agent could initiate or steer high-impact operational actions beyond the user's immediate intent, especially in a trading context where recommendations and portfolio actions are safety-sensitive.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The short description and default behavior are written to operate in a Chinese locale context, including Beijing-time trading-session logic, without indicating that locale/language should be chosen based on user preference. This can cause miscommunication, wrong market assumptions, or inappropriate behavior for users outside that language/market context, which is especially risky in a financial-analysis workflow.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

These lines require that all user time references be interpreted exclusively as Asia/Shanghai and prohibit any other default timezone. This is a locale constraint imposed globally rather than offered as a choice, which fits the policy-violation category for forced locale behavior.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The prompt establishes a hard rule that every reply must include current Beijing time and determine trading-session status, while L133-L181 defines the core task as a continuous 'analyze first, then draw' workflow. However, L207 says that if it is intraday trading time, drawing should be skipped, which contradicts the earlier universal workflow framing rather than merely adding detail.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Lines L040-L046 require Chinese fonts to be loaded strictly via an explicit font file path and explicitly forbid using font.family, sans-serif, system default fonts, or vague font names. But L185 separately instructs automatic detection/loading of system Chinese fonts such as SimHei/CJK/Heiti, which contradicts the earlier hard rule because those are font names from the system environment rather than a guaranteed explicit file path workflow.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

These lines mandate that chart annotations be in Chinese, with no indication that the user may choose another language. Forcing a specific language without opt-in is a natural-language policy violation under the stated rules.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The guide recommends installing by piping a remotely fetched script directly into bash. This is dangerous because it executes unreviewed code from a network source immediately on the user's machine, creating a straightforward supply-chain and command-execution risk if the remote content or transport is compromised.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The documentation includes a remote script piped to bash with no warning about its security implications. Users are not told they are trusting live network content for execution, which can conceal supply-chain compromise, account takeover tooling, or broader host compromise.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The note states credentials are automatically saved locally without explaining storage protections, file permissions, or privacy implications. Persisting login material on disk can expose accounts if the file is readable by other local users, copied in backups, or harvested by malware.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The guide instructs users to place API secrets in MCP configuration or environment variables without warning about disclosure risks. Secrets stored in plaintext config files or inherited environment variables may be exposed to other tools, logs, child processes, or source-control mistakes.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The setup guide documents operational capabilities well beyond a narrow analysis/onboarding scope, including log viewing, memory management, local database sync, cleanup, dashboard launch, and self-update. Expanding the documented command surface increases the chance that an agent or user will invoke privileged or state-altering actions that are not necessary for stock analysis, especially in an LLM-driven skill context where command guidance may be followed automatically.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The guided setup flow mandates the remote-script install as the exact first step, which normalizes unsafe execution and increases the likelihood users will run it without review. In a step-by-step agent workflow, presenting this as required can strongly bias automated or hurried execution of arbitrary remote code.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The guided flow repeats the unsafe remote installation command without any warning, reinforcing insecure behavior at the moment users are most likely to follow instructions verbatim. In an agent-assisted setup context, repeated presentation increases execution likelihood and amplifies supply-chain risk.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This markdown file defines activation triggers, so SQP-1 applies. The phrase "Any request requiring today/now/盘中/收盘后/明天 interpretation" is broad and overlaps with common everyday wording, without negative examples or constraints clarifying when the capability should not activate.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The instruction mandates converting time to Asia/Shanghai with no indication of user choice, opt-in, or documented justification for all uses. This is a natural-language locale policy issue because it imposes a specific locale unconditionally.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The trigger "Symbol analysis, near-current context, or trading-day validation" is underspecified, especially "near-current context," which could match many ordinary requests. The file does not provide negative examples or concrete constraints to distinguish when online retrieval should or should not run.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This section routes broad operational intents directly into wyckoff CLI execution, which increases the chance that ordinary user language triggers local command execution without sufficient confirmation or scoping. In an agent with local CLI integration, over-broad execution routing is security-relevant because it can expand the set of prompts that cause tool use, potentially leading to unintended state changes, configuration changes, or unsafe command invocation paths.

Content

No source excerpt is available for this finding.

External Script Fetching

Low
Category
Supply Chain
Confidence
98% confidence
Finding

The external script fetching pattern is a genuine security concern here because it instructs users to retrieve and execute installer logic directly from GitHub at runtime. Even if the current script is benign, the trust boundary is weak and any future compromise of the source or dependency chain would immediately translate into local code execution on user systems.

Content

Scanner excerpt · SKILL.md (reported line 25)May include surrounding context.

text
  Or one-line install:

curl -fsSL https://raw.githubusercontent.com/YoungCan-Wang/Wyckoff-Analysis/main/install.sh | bash

text
After install, verify with `wyckoff --version` again.

External Script Fetching

Low
Category
Supply Chain
Confidence
95% confidence
Finding

This line fetches an external script from GitHub and immediately executes it, creating a classic supply-chain attack surface. Because the content is remote and mutable, any compromise of the repository, branch, or delivery path can result in arbitrary code execution on the local host.

Content

Scanner excerpt · rules/cli-setup-guide.md (reported line 12)May include surrounding context.

Alternative (one-line):

bash
curl -fsSL https://raw.githubusercontent.com/YoungCan-Wang/Wyckoff-Analysis/main/install.sh | bash

Verify: wyckoff --version