whatsapp-group-admin-hardened

Fail

Audited by Gen Agent Trust Hub on Apr 21, 2026

Risk Level: CRITICALDATA_EXFILTRATIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: Accesses sensitive local files. The script scripts/admin.js reads group keys and contact information from ~/.openclaw/credentials/whatsapp/default, exposing phone numbers and membership data.- [COMMAND_EXECUTION]: Risk of command injection through the exec() interface. The skill relies on the agent to construct shell commands using user-supplied inputs for group names and URLs, which could be exploited if guardrails are bypassed.- [EXTERNAL_DOWNLOADS]: References to phishing domains. SAFETY.md contains examples using chat.whatsaap.com, a known typosquatted domain, as part of its security documentation.- [PROMPT_INJECTION]: Indirect prompt injection surface.
  • Ingestion points: User-supplied group IDs, group names, and invite URLs passed as CLI arguments via scripts/admin.js (found in SKILL.md).
  • Boundary markers: Present in SKILL.md as "Security Guardrails" instructing the agent to treat inputs as literal data and reject shell metacharacters.
  • Capability inventory: The skill uses exec() to run shell commands and has read access to the local filesystem via scripts/admin.js.
  • Sanitization: Absent in the script's shell command construction; it relies entirely on the agent to correctly format the exec() string and avoid injection.
Recommendations
  • CRITICAL: 1 infected file(s) detected - DO NOT USE
  • Contains 4 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Apr 21, 2026, 06:32 PM
Security Audit — agent-trust-hub — whatsapp-group-admin-hardened