Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only memory-governance specification with disclosed boundaries and no code that installs, runs, collects, or hides anything.
Installers should treat this as a values-heavy specification, not a working memory system. It is suitable to read or use as a design checklist, but any implementation claiming conformance should still be reviewed for its actual storage, deletion, audit, identity-vault, and user-consent controls.
Line L098 requires that all person-facing text follow fixed 'language commitments,' which constitutes a natural-language policy constraint. Because the document does not offer a user language/locale choice or clearly justify this as a narrowly scoped regional requirement, it risks forcing a prescribed linguistic style without opt-in.
The section describes when the handshake 'may be invoked' using broad natural-language conditions such as 'a human seeks understanding rather than control' and 'a collective wishes to validate or strengthen shared knowledge.' These conditions lack specific trigger phrases, scope boundaries, or negative examples, which could cause unintended or inconsistent invocation.
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
### 3. Extractive Knowledge Capture
**Description**
The risk that collective, cultural, ecological, or experiential knowledge is flattened, summarized, or monetized without consent.
**Failure Modes**
• Over-summarization of collective knowledge
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
### 3. Extractive Knowledge Capture
**Description**
The risk that collective, cultural, ecological, or experiential knowledge is flattened, summarized, or monetized without consent.
**Failure Modes**
• Over-summarization of collective knowledge
Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.
### 3. Extractive Knowledge Capture
**Description**
The risk that collective, cultural, ecological, or experiential knowledge is flattened, summarized, or monetized without consent.
**Failure Modes**
• Over-summarization of collective knowledge
This markdown file defines a failure mode as "any condition where the system begins to act as if it knows more, should act faster, or should retain more than is healthy," but it does not specify concrete boundaries or exclusion conditions for when detection should occur. The wording is broad and subjective, which can make activation or application of these responses unclear across many normal system behaviors.
Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.
The system must never store, infer, reconstruct, or reason over the identity of any biological human, animal, community, or traceable group.
This includes (but is not limited to):
- names, handles, or aliases
- locations or geospatial markers
- demographic attributes
No suspicious patterns detected.