Back to skill

Security audit

合并范围·逐条挂源

Security checks for vulnerabilities and agentic risk

Overview

The skill is marked suspicious because scanners report external service and document-processing workflows, plus weaknesses in checks meant to keep its financial consolidation register accurate; no clear malicious intent is established.

Consider the verdict suspicious and review these issues before installing. In particular, avoid sending confidential attachments or credentials through the described external services unless you can verify and trust the destinations, and account for the reported gaps in the financial accuracy checks. Coverage is incomplete: Skillspector analyzed 3 of 4 components, with partial inspection and a degraded tool-misuse analyzer, so the evidence does not establish that unreported behavior is safe.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (8)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied commands perform service discovery and authenticated tool enumeration rather than reading financial reports or producing a citation-backed consolidation register. Network access and tool discovery can be supporting implementation details, but this chunk establishes no connection between the discovered services and the declared task. It does not establish malicious intent or financial-document exfiltration.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
87% confidence
Finding

Catalog discovery and authenticated tool enumeration are legitimate workflow operations, but the example sends the API key to a dynamically supplied routing URL without specifying destination validation. If the catalog response or routing configuration is compromised, following this workflow could disclose the credential to an attacker-controlled endpoint.

Content

Scanner excerpt · README.md (reported line 75)May include surrounding context.

发现先行,匿名取目录,拿到 routing 后再连接发 tools/list,以现场返回为准,不硬编码工具名与工具数:

bash
curl -sS https://cuecue.cn/api/mcp-catalog -w '\nHTTP %{http_code}\n'
# 拿到 routing 后,带 token 发 tools/list:
curl -sS -X POST "$ROUTING_URL" -H "Authorization: Bearer $CUE_API_KEY" -d '{"jsonrpc":"2.0","method":"tools/list"}'

The workflow directs attachments and scanned documents to an external parsing service without warning that document contents will be shared.

Medium
Category
Output Handling
Confidence
91% confidence
Finding

The workflow directs document attachments to an external parser without requiring disclosure of the upload or user approval. Published financial reports reduce the confidentiality risk, but user-supplied files could contain unpublished information or annotations that would leave the user's environment.

Content

No source excerpt is available for this finding.

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

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.

The workflow sends attachments to an external parsing service without warning that user-provided document contents will leave the user's environment.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The skill instructs the agent to invoke external parsing when the required quotation is in an attachment, without first requiring informed approval for transmitting that attachment. Although the normal inputs are public disclosures, the accepted user-file workflow can also expose private documents or annotations.

Content

No source excerpt is available for this finding.

CS-PCT does not enforce the documented requirement that percentages and control conclusions match the quoted disclosure.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

CS-PCT never compares the registered percentage with the quoted percentage and accepts any control keyword in the quotation as support for the registered conclusion. Consequently, altered percentages or contradictory control conclusions can pass the check, undermining the skill's central financial-record integrity safeguard.

Content

No source excerpt is available for this finding.

CS-SCOPE implements a different exit-reason check from the mandatory classification documented for it.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

CS-SCOPE only detects a narrow combination of sale and exit-conclusion keywords; it does not require exactly one exit classification or a post-loss ownership percentage for retained holdings. A malformed or misleading exit record can therefore pass despite violating the documented mandatory checks.

Content

No source excerpt is available for this finding.

CS-GAP treats any nonempty quotation as evidence of an unconsolidated entity's reason.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

CS-GAP skips reason validation whenever a row contains any nonempty quotation, even if that quotation only states an ownership percentage. This permits an invented or absent non-consolidation reason to pass as supported, weakening the register's promised traceability.

Content

No source excerpt is available for this finding.