Sentient Memory

Manages cryptographically secured, consent-based relational memory with variance preservation, privacy safeguards, and adaptive dormancy under adversarial or...

This package is a normative design and operator handbook — not an executable plugin. It does not ask for keys or install software, so installing the skill itself does not expose credentials. Before adopting or implementing its recommendations, verify provenance and author identity, require independent security and cryptographic audits of any implementation (especially shard placement, key management, and revocation), and insist on running the listed conformance tests in a safe environment. Be cautious if someone later supplies code or an installer that claims to implement this spec: do not provide any secrets or grant persistent privileges until that implementation has been reviewed and its network/storage endpoints, key handling, and break-glass mechanisms are audited. If you want a lower-risk path, keep this as a read-only policy document rather than wiring it into live agent behavior until vetted.

Type: OpenClaw Skill Name: sentient-memory Version: 1.0.0 The OpenClaw AgentSkills skill bundle describes a comprehensive 'Sentient Memory & Deep Seed Protocol' focused on privacy, security, and decentralized governance. All files, including SKILL.md and other documentation, outline principles, parameters, and operational procedures for an AI agent to manage memory in a highly secure, consent-governed, and privacy-preserving manner. There are no instructions for malicious execution, data exfiltration, persistence, prompt injection, or any other harmful behaviors. The content consistently emphasizes safeguards against adversarial attacks, data concentration, and unauthorized access, indicating a clear intent for robust security and privacy.