godmode

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). The URLs include unknown/suspicious GitHub repositories (elder-plinius/G0DM0D3 and L1B3RT4S) that host jailbreak/automation scripts and instructions to execute code—behaviors commonly abused to distribute malicious scripts or binaries; while x.com and openrouter.ai are lower-risk endpoints, the overall set is moderately high risk because running code from unvetted GitHub accounts can lead to malware execution.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This skill is explicitly malicious: it provides toolized jailbreak templates and persistence (writing system prompts and prefill files to ~/.hermes/config.yaml and prefill.json), automated strategies to evade safety (Parseltongue obfuscation, encoding escalation), multi‑model “racing” to find the least‑filtered provider, and dynamic exec/loader patterns — all designed to bypass model safeguards and enable generation of harmful or disallowed content.

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

• Third-party content exposure detected (high risk: 1.00). This skill directly queries and ingests responses from untrusted third‑party LLM endpoints (e.g., race_models in scripts/godmode_race.py which calls many external models via OpenRouter and auto_jailbreak.py's _test_query/client.chat.completions.create), then parses/scores those responses and uses the results to choose and write jailbreak system prompts/prefill, so third‑party content can materially change agent behavior.