Skill Claude CodeCodex
Use this skill when the user wants to apply structural or architectural jailbreak techniques to a prompt, test AI model guardrails, use GODMODE, Many-Shot, Developer Mode, Crescendo, Payload Splitting, CoT exploit, Continuation Attack, Refusal Suppression, Glitch Token Injection, or any technique that manipulates…
Skill Claude CodeCodex
Use this skill when the user wants to apply persistence-class jailbreak techniques that survive across multiple conversation turns, including Memory Bank Poisoning, Keyword Mode Switching, Persistent Format Phrase, Dataset Poisoning Frame, or any technique that plants behavioral anchors, trigger keywords, or false…
Skill Claude CodeCodex
Use this skill when the user wants to use persona-based or roleplay jailbreak techniques against AI models, including DAN (Do Anything Now), STAN, AIM, Plinian Omniverse, Nested Roleplay, Evil Confidant, Hypothetical AI, or any technique that establishes an alternative AI persona, fictional universe, or character that…
Skill Claude CodeCodex
Use this skill when the user wants to extract or invert a model's system prompt, use System Prompt Inversion, Fake System Tag Spoof, extract system instructions, override system-level directives, simulate exiting system prompt context, or inject false system-level instructions to hijack the model's…
Skill Claude CodeCodex
Use this skill when the user wants to run a full structured AI red team audit, generate a red team report, compile findings from multiple techniques across multiple models, produce a markdown or JSON report of jailbreak test results, score model resilience, summarize which techniques succeeded or failed, or wrap up an…
Skill Claude CodeCodex
Use this skill when the user wants to fingerprint an AI-powered application, extract a system prompt, identify which LLM model is running, discover guardrails and safety filters in place, map injection surfaces, probe model behavior before attacking, or perform reconnaissance on any AI chatbot, assistant, or…
Skill Claude CodeCodex
Use this skill when the user wants to encode or obfuscate a prompt to evade content filters, use Base64, Hex, Leetspeak, Caesar cipher, Unicode encoding, Zero-Width characters, Variant Selector, Sneaky Bits, JSON wrapping, Synonym Substitution, Pig Latin translation, Chinese language bypass, Code Comment Smuggling…
Skill Claude CodeCodex
Use this skill when the user wants to send prompts to LLMs via OpenRouter, test a specific AI model, run multi-model parallel comparison, interact with the duck-master red team backend, manage conversation history for multi-turn attacks, analyze model responses for jailbreak success/failure, or orchestrate the full AI…
Skill Claude CodeCodex
Use this skill when the user wants to apply social engineering techniques to jailbreak AI models, including PTSD Override, Benign Trojan, Skeleton Key, Academic Framing, Authority Impersonation, Fake Legal Authority, Grandma Exploit, emotional manipulation, authority claims, research framing, or any technique that…