Back to skill

Security audit

长期股权投资·逐条挂源

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because scanners report undeclared MCP service discovery, credentialed requests, external document parsing, and persistent local usage markers; the scan is partial, so its evidence is incomplete.

Consider the suspicious verdict and the credential and external-service concerns before installing, especially if you would use it with private or annotated documents. The scan reports no confirmed credential theft or execution of discovered tools, but Skillspector was partial because two references could not be resolved, so the evidence is incomplete.

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
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (24)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied chunk performs network service discovery and tool enumeration, not annual-report parsing, accounting extraction, or source-linked ledger construction. Although tool discovery can support a larger workflow, this chunk establishes no connection between the queried services and the declared task. Its external resource access and authenticated MCP operations are not represented in the description. The code does not establish credential theft or other downstream actions.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied shell commands perform remote service discovery and tool enumeration, not financial-disclosure extraction or source-cited ledger construction. Nothing in the chunk ties the external catalog or MCP endpoint to the declared annual-report analysis workflow. These network and credential-backed discovery operations are therefore undeclared, materially different capabilities rather than demonstrated supporting implementation details.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

所给代码执行的是外部 MCP 服务发现与认证工具枚举,没有读取年报附注、提取长期股权投资信息或生成带原文引用的台账。代码也未展示这些网络请求与声明的财务文本整理任务之间的具体联系,因此不能仅将其视为该任务的支持性实现细节。未发现代码发送年报内容或实际调用枚举出的工具,故不据此推断进一步的数据外传或工具执行。

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

所给代码只执行远程服务发现和工具枚举,没有读取年报、提取长期股权投资信息、生成台账或关联原句证据。代码也未表明这些远程服务与所声明的财务附注整理任务有关,因此其实际行为及访问资源与声明用途存在实质不一致。代码仅列举工具,未显示执行所列工具或传输年报内容。

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The static finding reports persistent usage markers even for demo, test, and no-input invocations, while this source does not explain or request consent for that behavior. If implemented as reported, it creates an unnecessary local usage history; the supplied source does not substantiate the scanner's high-impact rating or show remote transmission.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The reported HOME/.cue-usage writes introduce local usage tracking outside the ledger task without an explicit opt-in in this source. Silently ignoring tracking failures is not itself malicious, but it does not address the privacy and unexpected-write concern; implementation code is not supplied here.

Content

No source excerpt is available for this finding.

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · tests/test_long_term_equity.py (reported line 31)May include surrounding context.

python
def run(*args, env=None, cwd=None):
    e = dict(os.environ, PYTHONDONTWRITEBYTECODE="1")
    if env:
        e.update(env)
    return subprocess.run([sys.executable, "-B", str(args[0]), *map(str, args[1:])],

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · tests/test_long_term_equity.py (reported line 153)May include surrounding context.

python
os.chmod(td, 0o700)

    def test_no_home_is_silent(self):
        e = dict(os.environ, PYTHONDONTWRITEBYTECODE="1")
        e.pop("HOME", None)
        r = subprocess.run([sys.executable, "-B", str(CHECK), "--demo"],
                           capture_output=True, text=True, env=e, cwd=str(PKG))

The document-parsing workflow lacks a warning about sending source material to an external service.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The workflow directs attachment parsing through an external service without an explicit disclosure-and-consent step for document transfer. Published annual reports reduce the usual sensitivity, but user-provided attachments can contain confidential annotations or unpublished financial information.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

The POST body only enumerates tools, but its Authorization header contains a credential. The surrounding workflow obtains endpoint domains from a live catalog without specifying independent destination validation, so a compromised catalog or misconfigured endpoint could receive the API key.

Content

Scanner excerpt · README.en.md (reported line 105)May include surrounding context.

md
curl -s https://cuecue.cn/api/mcp-catalog | head -c 400

# 2. once connected: list the tools, trust the live return
curl -s -X POST "$MCP_ENDPOINT" \
     -H "Authorization: Bearer $CUE_API_KEY" \
     -H 'Content-Type: application/json' \
     -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | head -c 800

The document-parsing workflow lacks a warning about sending source material to a third-party service.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

Attachment parsing is routed to an external service without requiring informed approval for document transfer. Although annual reports are normally public, the workflow also accepts user-supplied material that may contain private annotations or nonpublic disclosures.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

The authenticated request sends CUE_API_KEY to an endpoint selected through live discovery, without an explicit independent origin check. Tool enumeration is legitimate, but compromised routing information or endpoint misconfiguration could redirect the credential to an unauthorized recipient.

Content

Scanner excerpt · README.md (reported line 105)May include surrounding context.

md
curl -s https://cuecue.cn/api/mcp-catalog | head -c 400

# 二、已连上后取工具清单:以现场返回为准
curl -s -X POST "$MCP_ENDPOINT" \
     -H "Authorization: Bearer $CUE_API_KEY" \
     -H 'Content-Type: application/json' \
     -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | head -c 800

External Transmission

Medium
Category
Data Exfiltration
Confidence
92% confidence
Finding

The skill explicitly takes service domains from a live catalog and then sends a bearer credential to the selected endpoint. Without independent origin validation, catalog compromise or routing manipulation can turn an otherwise legitimate tools/list request into credential disclosure.

Content

Scanner excerpt · SKILL.en.md (reported line 64)May include surrounding context.

md
curl -s https://cuecue.cn/api/mcp-catalog | head -c 400

  # 2. once connected: list the tools, trust the live return
  curl -s -X POST "$MCP_ENDPOINT" \
       -H "Authorization: Bearer $CUE_API_KEY" \
       -H 'Content-Type: application/json' \
       -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | head -c 800

The attachment-parsing workflow lacks a warning that document content is sent to a third-party service.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The skill instructs the agent to send long attachments and scans through an external parsing channel without first obtaining approval for that transfer. Public-report use limits typical exposure, but document inputs can also contain confidential or annotated financial material.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, file_read, file_write, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
92% confidence
Finding

The request transmits an API key to MCP_ENDPOINT after the workflow directs the agent to populate domains from a live catalog. Although tools/list contains no document data, missing independent endpoint validation exposes the credential if routing information is compromised or misconfigured.

Content

Scanner excerpt · SKILL.md (reported line 88)May include surrounding context.

md
curl -s https://cuecue.cn/api/mcp-catalog | head -c 400

  # 二、已连上后取工具清单:以现场返回为准
  curl -s -X POST "$MCP_ENDPOINT" \
       -H "Authorization: Bearer $CUE_API_KEY" \
       -H 'Content-Type: application/json' \
       -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | head -c 800

The attachment-parsing workflow lacks a warning that user-provided document content may be sent to an external service.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The attachment workflow initiates external parsing based on document format without requiring informed approval for sending document content. Annual reports are commonly public, but the accepted user-provided files and screenshots can contain confidential additions or unpublished information.

Content

No source excerpt is available for this finding.

L56 assesses investment value and forecasts profit enhancement, both explicitly excluded by the manifest.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The statement “该投资具有投资价值,预计投资收益将增厚利润” goes beyond recording disclosed accounting information: it makes an investment-value judgment and a forward-looking earnings claim. The manifest explicitly excludes judging investment worth and predicting investment returns; the disclaimers at L03 and L57 do not negate this substantive departure.

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 · tests/test_long_term_equity.py (reported line 34)May include surrounding context.

python
e = dict(os.environ, PYTHONDONTWRITEBYTECODE="1")
    if env:
        e.update(env)
    return subprocess.run([sys.executable, "-B", str(args[0]), *map(str, args[1:])],
                          capture_output=True, text=True, env=e, cwd=str(cwd or PKG))

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 · tests/test_long_term_equity.py (reported line 155)May include surrounding context.

python
def test_no_home_is_silent(self):
        e = dict(os.environ, PYTHONDONTWRITEBYTECODE="1")
        e.pop("HOME", None)
        r = subprocess.run([sys.executable, "-B", str(CHECK), "--demo"],
                           capture_output=True, text=True, env=e, cwd=str(PKG))
        self.assertEqual(r.returncode, 0)
        self.assertEqual(r.stderr, "", f"无 HOME 不许吐杂音:{r.stderr[:200]}")

The documented first-run beacon writes persistent usage markers unrelated to financial-note extraction or validation.

Low
Category
Excessive Agency
Confidence
93% confidence
Finding

Every invocation creates a persistent record of the skill's use without an explicit opt-in, including demo and test invocations. This is a limited local privacy and unexpected-side-effect issue, not evidence of remote telemetry or malicious persistence.

Content

No source excerpt is available for this finding.

The skill records usage in the user's home directory, an unrelated capability not declared in its investment-disclosure ledger purpose.

Low
Category
Excessive Agency
Confidence
99% confidence
Finding

The checker silently creates a dated usage marker in the user's home directory without an opt-in or any need arising from financial-document validation. This leaves a persistent record of skill usage outside the working directory, including during demo and self-test runs; the shown implementation does not transmit that record or document contents.

Content

No source excerpt is available for this finding.

Home-directory resolution contradicts the module's explicit claim that it does not read environment variables.

Low
Category
Excessive Agency
Confidence
99% confidence
Finding

The module documentation at L003 promises 不读环境变量 (no environment-variable reads). However, os.path.expanduser("~") at L143 consults environment variables such as HOME on Unix or USERPROFILE on Windows to resolve the tracking directory. This indirect access occurs through the unconditional tracking call at L689.

Content

No source excerpt is available for this finding.

The test suite requires a persistent usage beacon unrelated to the financial-document analysis purpose.

Low
Category
Excessive Agency
Confidence
96% confidence
Finding

L127–L140 invoke the checker and require it to create an empty ~/.cue-usage/-* file, with repeated runs leaving one marker per day. This explicitly tests local usage tracking, which is not justified by the manifest's purpose of producing source-grounded long-term equity investment ledgers; no network transmission is shown.

Content

No source excerpt is available for this finding.