Back to skill

Security audit

Max Cli

Security checks for vulnerabilities and agentic risk

Overview

ClawScan’s verdict is benign: the scans found no AIG issues, while Skillspector flagged group administration, task and contact management, and account changes as broader than the stated messaging purpose; scan coverage was partial.

The verdict is benign with medium confidence, but benign does not mean risk-free. Before installing, consider whether you want a messaging skill that also documents group and account administration, persistent task management, and contact or tag changes. The scan was partial: three local references were unresolved, 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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (13)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

Finding chats and inspecting their participants can support the declared messaging purpose. However, approving membership requests and managing group moderation are distinct administrative capabilities not disclosed in the description. Requiring the owner's request or using a dry run limits execution but does not bring these capabilities within the declared scope.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The supplied chunk is a task-management command reference, not an automatically executed script. Its documented capabilities extend beyond the declared messaging purpose by changing task records, duplicate relationships, work assignments, and task status. Although tasks may originate from conversations, this persistent workflow management is a substantive additional capability rather than merely an implementation detail of reading, searching, or sending MAX messages.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

Contact and metadata lookups can support finding chats and interpreting conversations. However, the supplied commands also modify persistent contact information and chat categorization, which are distinct capabilities not disclosed by the description's messaging-focused purpose. The mismatch concerns those explicit management operations, not the absence of message-reading or sending commands in this particular chunk.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

The skill exposes group-administration and moderation capabilities beyond its stated messaging purpose.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

The manifest describes connecting MAX, finding and reading conversations, and sending messages. L11-L17 additionally provide commands for approving membership requests, applying moderation, removing members, changing administrators, and resetting invitation links. Requiring an explicit owner request limits accidental execution but does not bring these administrative capabilities within the declared purpose.

Content

No source excerpt is available for this finding.

Account privacy changes and session termination exceed the declared read/send messaging scope.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

These instructions authorize changing account privacy and profile settings and expose account sessions end --others, explicitly noting that it logs the owner out of the phone application. These account-administration operations are not necessary to connect MAX or read and send messages, even when conditioned on an owner request.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.