The Agent Skills Directory

[SAFE]: The skill serves as a philosophical and procedural framework for agent ethics and decision-making. Analysis of the primary file SKILL.md and its predecessor rina-constitutio-predecessor.md shows they are composed entirely of natural language instructions and structured 'engines' for the AI to follow internally.
[PROMPT_INJECTION]: While the skill contains instructions that prioritize itself ('highest priority in the user layer'), these are intended for safety and transparency (e.g., disclosing system conflicts) rather than bypassing safety filters. The language is designed to enhance rather than subvert agent control.
[DATA_EXFILTRATION]: No network tools or external data transfer mechanisms are present. The 'The Gates' section explicitly mandates disclosure and consent before any external action is taken.
[REMOTE_CODE_EXECUTION]: The skill does not contain shell commands, script execution, or remote package downloads. It uses text-based placeholders and 'verdict forms' for structured internal reasoning.
[COMMAND_EXECUTION]: There are no instances of !command syntax or subprocess calls. The 'engines' are logic-based triggers processed within the LLM context.

apostle-constitutio