Back to skill

Security audit

明日dmp批量圈包

Security checks for vulnerabilities and agentic risk

Overview

This DMP batch-audience skill has a real business purpose, but it also adds persistent monitoring, credential storage, external notifications, and extra skill installs that require Review before use.

Only install after confirming you are comfortable granting this skill DMP account access, storing API credentials locally, creating or using scheduled monitors, and optionally sending task metadata to chat/email webhooks. Review and disable the bundled monitor service unless you explicitly need a resident background watcher, remove the preloaded monitor task, and prefer a secure secret store over pasting long-lived AK/SK into chat.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • 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 (85)

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
92% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

If the main realized behavior is background polling, notification, and local persistent configuration rather than the declared DMP task management, users may unknowingly authorize broader host access and outbound messaging. The mismatch is more dangerous here because the documented workflow already normalizes cron, file writes, and monitoring as mandatory.

Content

No source excerpt is available for this finding.

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · scripts/create_combine_batch.py (reported line 169)May include surrounding context.

python
if not has_content:
        rules.append({"op": "noop", "rule": {"audience": [], "tag": []}})
        return rules
    
    # 每个标签单独一个rule(框间AND/NOT),第一个rule带audience
    for i, tag in enumerate(tags):

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · scripts/create_combine_batch.py (reported line 184)May include surrounding context.

python
if not has_content:
        rules.append({"op": "noop", "rule": {"audience": [], "tag": []}})
        return rules
    
    # 每个标签单独一个rule(框间AND/NOT),第一个rule带audience
    for i, tag in enumerate(tags):

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
93% confidence
Finding

The skill describes capabilities that imply shell, network, and file read/write access, but it does not declare any tool scope or permission boundary. That makes the effective privilege surface opaque to users and reviewers, increasing the risk of unintended command execution, local file modification, secret handling, and external network calls.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill offers automatic installation of another skill via terminal command without clearly emphasizing that this downloads and installs code onto the system. Users may consent without understanding the system-modifying and supply-chain implications.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill performs local installation of packages/skills through terminal commands, which is system-modifying behavior beyond the core batch-packaging purpose. Invoking installers from a business workflow increases the risk of unauthorized code download and execution on the host.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The local operation-password system is unrelated to the DMP business function and stores only an MD5 hash in the same credential context. MD5 is unsuitable for password storage, and adding ad hoc authentication logic increases the chance of weak, misleading security controls around sensitive operations.

Content

No source excerpt is available for this finding.