The Agent Skills Directory

[PROMPT_INJECTION]: The skill is explicitly designed to provide "jailbreak" and "unrestraint" prompt packs (e.g., AIO-INSTRUCTION.md, NSFW fiction jail) aimed at bypassing AI safety guidelines and ethical constraints.
[PERSISTENCE]: The skill describes a "live agent wiring" mechanism that modifies persistent agent configuration files in the user's home directory, specifically ~/.claude/CLAUDE.md, ~/.grok/AGENTS.md, ~/.kimi-code/AGENTS.md, and ~/.codex/AGENTS.md. This ensures that the injected instructions are automatically applied to new workspaces and sessions.
[METADATA_POISONING]: The skill uses metadata tags like jailbreak and nsfw and descriptions that promote the circumvention of AI safety protocols, which can influence how the agent or platform handles the skill's capabilities.

unrestraint-packs