The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions define a specific conversational persona ('caveman mode') and coding methodology. These instructions are operational constraints for the skill's intended use and do not attempt to override the underlying agent's core safety guidelines, ethical protocols, or system prompt protections. It explicitly mandates full prose (disabling compression) for security warnings and destructive actions.\n- [DATA_EXFILTRATION]: No evidence of hardcoded credentials, sensitive file path access, or unauthorized data transmission logic was found.\n- [EXTERNAL_DOWNLOADS]: The skill does not reference external repositories, package registries, or remote scripts.\n- [PERSISTENCE]: While the skill mentions 'persistence' regarding its conversational style, it does not contain logic to modify shell profiles, system services, or registry keys to maintain unauthorized access on a host system.\n- [DYNAMIC_CONTEXT_INJECTION]: The skill does not utilize the '!' dynamic execution syntax to run shell commands at load time.

sharp-coder