Back to skill

Security audit

限售解禁压力·逐条挂源

Security checks for vulnerabilities and agentic risk

Overview

The scanners report a benign verdict with medium confidence: the skill’s main concerns are undisclosed remote tool discovery, inconsistencies in its financial-data rules, and an unpinned install command.

The verdict is benign, not risk-free, and scanner coverage is incomplete: 14 of 15 components were fully inspected, with one partially inspected and a static analysis limitation. Before installing, consider whether the external tool-discovery steps are necessary, whether you trust the unpinned install sources, and whether the percentage and cost-field rules match your needs.

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 (18)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The chunk implements remote service discovery, with no visible retrieval of stock-unlock announcements, calendar construction, source attribution, or execution of check_lockup.py. These network operations are not established as supporting the declared workflow. The explicit zero-network statement concerns the validator, so it does not necessarily prohibit all networking elsewhere; nevertheless, the supplied code's demonstrated behavior is undisclosed MCP discovery rather than the stated functionality.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The description promises that percentages are not calculated from share counts, but the operational instructions explicitly permit calculated percentages when accompanied by a methodology note. This weakens the advertised source-only integrity constraint and can allow derived financial figures into a report consumers expect to contain disclosed values; explicit labeling reduces, but does not eliminate, that risk.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The source-only percentage restriction in the description conflicts with the documented exception allowing self-calculation. This is a bounded financial-data integrity issue rather than evidence of malicious behavior, because the exception is disclosed and requires a methodology note.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

The observable behavior is outbound HTTP-based discovery of remote tools. No lockup-specific retrieval, calendar construction, source tracking, or validation appears in this chunk, and nothing establishes that the external catalog supports those tasks. The zero-network statement specifically describes check_lockup.py, so these requests do not by themselves prove that checker violates its promise; nevertheless, the supplied code introduces an undisclosed external discovery capability rather than demonstrating the declared workflow.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
88% confidence
Finding

Installation alone is a supporting implementation detail and is not the basis for this finding. The material discrepancy is the additional remote MCP discovery workflow, which is absent from the description and supplied instead of the advertised tracking or validation behavior. The zero-network wording specifically qualifies check_lockup.py, so this snippet does not prove that checker itself violates its guarantee; nevertheless, the supplied workflow introduces undisclosed external service access.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
86% confidence
Finding

Installing the named skill is ordinary supporting setup. However, the snippet additionally contacts an external service to obtain routing information and enumerates tools at a remote endpoint, introducing an undisclosed remote-tool dependency. The network-free claim specifically applies to check_lockup.py, so these commands do not establish that the checker itself violates that claim. The mismatch is the additional external discovery behavior, not merely the absence of tracking or validation code in this excerpt.

Content

No source excerpt is available for this finding.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 131)May include surrounding context.

md
bash scripts/fixtures/run_fixtures.sh # 离线自检,30 秒,零网络零计费(档数由末行现算报出,不写死)

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The installation command invokes an unpinned npm executable and installs skill content from a mutable repository reference. A compromised or malicious future release could execute with the installer's privileges or introduce adversarial skill instructions; this is a supply-chain exposure, not specifically an unversioned MCP server.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · README.en.md (reported line 61)May include surrounding context.

CATALOG="https://cuecue.cn/api/mcp-catalog" curl -s "$CATALOG" | head -c 2000

3. confirm the tools are present

curl -s -X POST "$ROUTING_URL"
-H 'Content-Type: application/json'
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

text

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · README.md (reported line 61)May include surrounding context.

CATALOG="https://cuecue.cn/api/mcp-catalog" curl -s "$CATALOG" | head -c 2000

3. 确认工具在场

curl -s -X POST "$ROUTING_URL"
-H 'Content-Type: application/json'
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

text

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.en.md (reported line 86)May include surrounding context.

CATALOG="https://cuecue.cn/api/mcp-catalog" curl -s "$CATALOG" | head -c 2000

take the group's routing address from the catalog, then confirm the tools are present

curl -s -X POST "$ROUTING_URL"
-H 'Content-Type: application/json'
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

text

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 102)May include surrounding context.

CATALOG="https://cuecue.cn/api/mcp-catalog" curl -s "$CATALOG" | head -c 2000

从目录里取本组的路由地址,再确认工具在场

curl -s -X POST "$ROUTING_URL"
-H 'Content-Type: application/json'
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

text

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The documented installation resolves an unpinned npm CLI and mutable skill repository content. Compromise of either distribution source could introduce code execution or malicious agent instructions under the user's privileges.

Content

No source excerpt is available for this finding.

The output contract mandates Chinese-language labels and markers without offering a language choice.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The skill requires verbatim Chinese header markers, Chinese field prefixes, and Chinese-only holder-type labels, regardless of the user's language. Although it covers A-share announcements, it does not explain why the generated report must retain these Chinese-language elements or offer localized alternatives.

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 validator accepts self-calculated percentages despite the manifest's explicit prohibition.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The validator accepts a self-calculated percentage whenever section six contains a methodology marker, and an explicit positive fixture preserves that behavior. This fails to enforce the manifest's source-only restriction, allowing derived financial figures to receive a clean validation result; the visible labeling and human-review requirements limit the impact.

Content

No source excerpt is available for this finding.

The fixture suite requires acceptance of self-calculated share percentages despite the manifest explicitly prohibiting that methodology.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The fixture explicitly requires acceptance of self-calculated percentages when their methodology is disclosed, contradicting the skill's stated prohibition on calculating those percentages. This weakens the financial-data integrity boundary by encoding prohibited behavior as a successful test outcome, although this source alone does not establish the checker's runtime behavior.

Content

No source excerpt is available for this finding.

Cost information is populated from the first available price despite documentation promising a placeholder.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

L009 explicitly states that cost ranges remain placeholders for manual completion. However, L127–L128 select the first nonempty price across all unlocking rounds, and L137 inserts it into the generic “第 N 轮成本区间” field at L040. This discards round-specific cost distinctions and presents a supplied price where the documented workflow requires an unfinished field.

Content

No source excerpt is available for this finding.