social-autoposter

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). These URLs include direct downloadable scripts/installer pages on non-official or ambiguous domains and a likely typosquatted domain (cl0ne.ai) which are common vectors for distributing malware (direct .sh installers and custom "install" pages are high-risk when from untrusted domains).

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This codebase contains multiple deliberate features for evading platform detection, creating/operating burner persona accounts, and accessing/exfiltrating sensitive session and machine-identifying data to a hosted service (s4l.ai), indicating high abuse risk.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). The required workflow for this skill (“Social Autoposter”) involves scanning and replying to responses on Reddit/X/LinkedIn/Moltbook, which necessarily ingests outsider-authored free text (e.g., their_content from inbound comments/replies) into the agent’s LLM context during drafting/replying.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly directs running npx/install-time commands that fetch and execute remote npm-hosted code (see https://www.npmjs.com/package/social-autoposter) and instructs installing Homebrew from https://brew.sh, so remote code is pulled and run during setup/runtime.