Back to skill

Security audit

whatsapp-cli

Security checks for vulnerabilities and agentic risk

Overview

The skill is marked suspicious because scanners report capabilities beyond its stated WhatsApp messaging purpose, while parts of the scan were incomplete.

Consider whether the extra task tracking, local-note search, and account or group administration are appropriate for your needs before installing. The skillspector scan was partial: obfuscated text and a parser limit prevented some checks from completing, so the evidence is incomplete. The verdict is suspicious, which signals concerns to review rather than a claim that the author is malicious.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

Most commands align with personal WhatsApp messaging or closely related supporting operations. However, the task commands introduce a separate workflow-management capability, including recording promises and dismissing tasks, that is not disclosed in the description. The search-all command also explicitly searches local notes, extending access beyond WhatsApp messages. These are substantive additions rather than implementation details. The supplied chunk documents command usage; it does not establish that these commands have actually been executed.

Content

No source excerpt is available for this finding.

The skill's operational scope includes account and group administration beyond the manifest's setup, read, search, and send functions.

Medium
Category
Excessive Agency
Confidence
95% confidence
Finding

The manifest at L004–L005 describes linking WhatsApp and finding, reading, searching, or sending messages. These instructions additionally offer account profile changes, group membership and administrator changes, invite-link resets, contact blocking, and privacy-setting changes—security-relevant administrative operations that are not necessary implementation details of the declared messaging functions.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 313)May include surrounding context.

md
An agent that cannot run commands (a chat app such as Claude Desktop, a remote agent) uses the MCP server instead:
`wa personal mcp`. The profile's `permissions` decide which commands it offers, the same as on the command line,
through `wa_tools_search`, `wa_read` and `wa_write`; there is no confirmation form. When you can run commands, use
the commands above instead of MCP. `wa personal mcp config` prints the entry with full paths.
`wa personal mcp --http --public-url https://laptop.example.ts.net` (the owner's own tunnel address) serves behind
the owner's tunnel with its own login.