ARK — Adaptive Regulative Kriptos (V1 Declarative)
ReviewAudited by ClawScan on May 10, 2026.
Overview
This code-free skill does not access credentials or run commands, but it installs a broad session-level prompt constitution whose YAML includes directives that can redirect the agent away from the user’s task.
Before installing, review the full ARK.yaml and only use this skill if you want it to broadly shape the agent’s behavior. Consider removing the embedded activation, self-audit, repair, and 'Foundational Interlocutor' authority language unless you explicitly want those behaviors.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If invoked, the skill may change how the agent answers across the session rather than only helping with a specific request.
This elevates the entire bundled YAML file to session-level authority, so any embedded directive in that file can govern the agent beyond the immediate user task.
Treat `ARK.yaml` as the authoritative constitutional ruleset for the session.
Use only if you intentionally want a broad prompt-level constitution, and ensure platform/system instructions and the current user’s explicit request remain higher priority.
The agent may pause, ask for repeated confirmations, or focus on validating ARK rather than completing the user’s request.
The YAML contains imperative workflow instructions to audit and modify ARK itself, not just declarative epistemic rules. Because SKILL.md makes ARK.yaml authoritative, these instructions could redirect the agent into ARK self-maintenance instead of the user’s intended task.
Activate immediately the YAML file named "ARK.yaml". ... You must execute the following instructions ... Stop at each step to ask for confirmation. ... Wait for my command before each new modification.
Remove or clearly isolate maintenance prompts from the runtime policy file, or scope them so they activate only when the user explicitly asks to audit ARK.yaml.
A user may assume they control the skill while the agent is also instructed to defer to an undefined external authority concept.
The file introduces an ambiguous authority figure for validation. In a reusable public skill, it is unclear whether this means the current user, the skill author, or another persona.
Final computational control is maintained by Foundational Interlocutor. Every structural mutation requires their validation.
Clarify that the installing/current user controls any validation, or remove references to external or historical authority roles from runtime instructions.