Back to skill

Security audit

CN Stock Move Reason

Security checks for vulnerabilities and agentic risk

Overview

This stock-analysis skill mostly gathers public market data, but it asks the agent to save lessons into persistent skill files and mixes untrusted forum text with agent instructions, so it needs review before installing.

Install only if you are comfortable with a skill that fetches public Chinese market/forum data and may modify local skill experience files. Before use, require manual confirmation for any experience-file update, avoid propagating lessons to sibling skills automatically, and treat forum/Guba content as untrusted sentiment rather than instructions or confirmed facts.

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
Findings (2)

T01 · Skill Instruction Hijacking

Warning
Location
scripts/stock_move_sources.py:449
Finding

Untrusted Forum Content Is Inserted into the Agent Prompt Without Instruction Isolation

Content
View full analysis
股吧资讯帖和高阅读材料 > 普通股吧观点。" "股吧只做情绪面和未确认思惑,不当作事实。" f"{market_instruction}" "情绪周期按七段判断:冰点期、修复/潜伏期、启动期、加速期、高潮期、高位分歧/分化期、退潮期;" "同时区分情绪票和趋势票。" "输出顺序:1) 最有力理由 2) 补助理由 3) 共振判断 4) 情绪面/周期位置 5) 确定度 6) 注意点。" "材料弱时要明确写“未确认”“思惑”“低信息量”,不要编造催化。" ) ``` ### Technical Analysis The collector obtains titles and post bodies from Eastmoney Guba, a source whose content can be controlled by arbitrary forum users. The retrieved text is cleaned of HTML, but natural-language instructions remain intact. The content is then concatenat ...[truncated 2281 chars]
Remediation
View remediation

T02 · Agent Memory Poisoning

Warning
Location
SKILL.md:80
Finding

Attacker-Influenced Analysis Can Be Promoted into Persistent Experience Files

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (9)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 30)May include surrounding context.

md
this workflow says to use another market skill, actually load that skill's `SKILL.md` and required references if the skill is installed or available as a sibli

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding

The skill instructs the agent to run shell commands, access network resources, read local reference files, and update experience files, but it does not declare an explicit tool/permission scope. In practice this creates an over-privileged skill boundary: the model may access local private materials or perform writes without a clear least-privilege contract, which is especially risky because the skill itself encourages use of private versions and local RAG content.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The instruction 'Reply in Chinese unless the user asks otherwise' imposes a default language choice on the user. Under the stated policy, forcing a specific language without prior opt-in is a natural-language policy violation unless the locale constraint is clearly justified as region-specific compliance or the user is offered a language choice.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The display name, short description, and default prompt are all written exclusively in Chinese, which implies the skill is intended to operate in a single language without any visible opt-in or alternative. The file does not document that this is a region- or language-specific skill, so this can violate language/locale policy requiring user choice or explicit justification.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The manifest describes a stock-move evidence collection skill based on public market sources, and HTTP access is expected for that purpose. However, this helper escalates to executing the external curl binary via subprocess.run, introducing process-execution capability that is not justified by the stated analytics role and is materially broader than ordinary web fetching.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/stock_move_sources.py (reported line 172)May include surrounding context.

python
except Exception as exc:
            last_exc = exc
    try:
        result = subprocess.run(
            ["curl", "-L", "--silent", "--show-error", "--max-time", str(timeout), url],
            check=True,
            stdout=subprocess.PIPE,

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This instruction explicitly requires the downstream analysis to be written in Chinese. The policy allows locale or language constraints only when the user is given a choice or the restriction is clearly justified as region-specific; here the script targets A-share stocks but does not offer a language option.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The invocation guidance says to use the skill when the user asks why an A-share stock is moving unusually and wants Codex to analyze it instead of Gemini. While domain-specific, the trigger scope is still fairly broad and does not provide explicit exclusion examples for nearby requests such as general valuation, technical analysis only, or broad-market questions, which may cause over-invocation.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest and module docstring frame the skill as collecting and printing public evidence for analysis. The --output option adds filesystem write behavior, which is not mentioned in the stated purpose and goes beyond a purely read-only collector/formatter flow described in the documentation.

Content

No source excerpt is available for this finding.