Security audit
Security checks for vulnerabilities and agentic risk
The skill is a markdown-only memory governance protocol, but it defines persistent memory and emergency audit behavior with some ambiguous triggers and an under-scoped session-log review path.
Use this only if you are deliberately building a persistent, consent-governed memory system. Before installing in a real environment, require measurable emergency triggers, explicit approval paths, scoped forensic access, and a replacement for blanket session-log review using minimized integrity proofs or time-bound audited access.
This markdown file includes an activation cue, "User asks for planning/risk/repair/execution/synthesis," that is very broad and lacks clear boundaries or exclusions. Several listed intents are common request types, so the trigger could overlap with ordinary conversation and cause unintended invocation.
The AEGIS trigger terms like extraction pressure, coercion attempt, narrative dominance anomaly, and variance compression are conceptually meaningful but operationally undefined. In a security-sensitive memory governance protocol, ambiguous triggers can be abused to justify unnecessary freezes, key rotations, or shard redistribution, or can fail to activate when genuinely needed because implementers interpret them inconsistently.
The protocol explicitly states that auditors should verify proofs and claims without raw logs or identity-linked content access, but the emergency recovery example later instructs checking all previous session logs for tamper. That contradiction can normalize broader log access during crises, creating a policy loophole that may expose sensitive data and undermine the stated privacy model.
No suspicious patterns detected.
