The Agent Skills Directory

[SAFE]: The skill implements a cognitive deliberation framework. It defines 12 thinking archetypes to simulate a multi-perspective discussion on user-provided topics. The logic is entirely instructional and does not involve any executable code or risky tool usage.\n- [DATA_EXPOSURE]: There are no attempts to access sensitive files (e.g., .ssh, .aws, .env), credentials, or system environment variables. Access is limited to standard Read/Write tools as specified in the frontmatter, with no instructions to abuse them for exfiltration.\n- [PROMPT_INJECTION]: The instructions do not contain patterns for bypassing safety filters, disregarding previous instructions, or extracting system prompts. The 'Phase 2' and 'Phase 3' instructions are standard formatting requirements for the deliberation process.\n- [REMOTE_CODE_EXECUTION]: The skill does not download external scripts, install packages, or use dynamic execution functions like eval() or exec(). There is no mention of remote servers for code execution.\n- [OBFUSCATION]: Analysis for Base64 encoding, zero-width characters, homoglyphs, and other encoding techniques returned no results. All content is in plain text and consistent with the stated academic/strategic purpose.

alterlab-consciousness-council