ARK — Adaptive Regulative Kriptos (V1 Declarative)

This skill is an instruction-only 'constitutional' layer that asks the agent to read and apply a local ARK.yaml ruleset; it does not request credentials, install software, or contact external endpoints. Before installing, consider: - Confirm you are comfortable with a session-level 'constitution' that will constrain agent outputs (confidence percentages, refusal behavior, ban on unrequested suggestions). This can change how the agent responds across tasks. - The ARK.yaml asks the agent to audit itself and produce proposed edits (a preface and single-item remediation proposals) but also states the AI must not write changes to the file and that the human must apply changes manually — ensure your agent runtime will not automatically persist or apply generated edits. - The YAML references many internal tracers and ILR IDs (conceptual labels). These are labels only; there is no code that implements them here. If other installed skills implement similarly named tracers, review for policy conflicts or duplicate naming. - Because the skill can influence reasoning norms, review the first few outputs carefully to verify it behaves as you expect (it may refuse or downgrade answers that it deems 'important' by its rules). - If you use autonomous agents, decide whether to allow this skill to be auto-invoked: it could change agent behavior across sessions. If you want to avoid unexpected cross-session behavior, do not grant it automatic invocation or monitor its first runs. Overall, the package is internally consistent with its stated purpose; the main caution is about the operational impact of adding a declarative constitutional layer to an agent and ensuring generated edit proposals are reviewed and not auto-applied.