twitter-digest

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). These URLs include a third‑party GitHub repo and raw.githubusercontent.com install scripts that the installer fetches/executes (curl|sh and git clone), which is a common malware distribution pattern and thus should be treated as suspicious compared with the other official X API and local callback URLs.

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

• Third-party content exposure detected (high risk: 0.85). Yes—runtime path scripts/run_daily_digest.py → digest_context.py reads API-derived text into digest-context.md (via digest-input.json), and that markdown is then used as the agent’s LLM input (ai_input_markdown), where the text originates from outsider-authored X posts/DM messages.

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

• Potentially malicious external URL detected (high risk: 1.00). The installer and README show the skill fetching and executing remote install code (curl | sh and curl -o ... && /bin/sh) from https://raw.githubusercontent.com/BofAI/skills/v1.5.13/twitter-digest/install.sh and the install script also clones and runs code from https://github.com/BofAI/skills.git (the installer constructs https://raw.githubusercontent.com/BofAI/skills/${TAG}/twitter-digest/install.sh and executes it), so remote content is fetched and executed at install/runtime.