Back to skill
Skillv9.0.0

ClawScan security

Llm Memory Interface Only · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignApr 16, 2026, 3:13 AM
Verdict
benign
Confidence
high
Model
gpt-5-mini
Summary
This package is an interface-only skill that only declares abstract Python interfaces and requests no credentials or installs; the contents are coherent with its description, with minor documentation-level inconsistencies to be aware of.
Guidance
This package appears to be a harmless, interface-only library. Before you install or run anything from the external implementation repository referenced in the README (https://cnb.cool/llm-memory-integrat/llm.git): 1) Inspect that implementation's code for network calls, subprocess usage, or credential consumption; 2) Prefer cloning from trusted/official hosts when possible; 3) If the implementation requires API keys (VectorConfig.api_key), only provide keys with appropriate least privilege and rotate them if needed; 4) Treat the cnb.cool repository as unverified until you audit it or obtain the implementation from a known source. If you only need the abstract interfaces and will implement them yourself, no additional risk is introduced by this package.

Review Dimensions

Purpose & Capability
okThe name/description say this is a pure interface package and the included Python files are abstract base classes and dataclasses defining Memory, Search, and Vector interfaces. No unrelated binaries, env vars, or capabilities are requested — the declared purpose matches the code.
Instruction Scope
noteSKILL.md consistently states this is an interface-only package with no network/file/subprocess usage, and runtime instructions do not ask the agent to read local secrets. However, the README suggests cloning an external private repository (https://cnb.cool/llm-memory-integrat/llm.git) to obtain implementations. That cloning step would be a network operation performed by the user (not the skill itself) and is optional, so the skill's runtime instructions remain limited; still, the documentation claim 'network: false' is slightly misleading because it points users to an external repo.
Install Mechanism
noteThere is no install spec and no code that would be executed at install time. The package.json references a repository URL hosted on cnb.cool (an external/unfamiliar domain). While the skill itself doesn't auto-download code, the README's guidance to git clone that URL means installing the implementation would involve fetching code from an unvetted host — review that implementation before running it.
Credentials
okThe skill declares no required environment variables or credentials. The VectorConfig dataclass includes optional fields like api_key and base_url, which is reasonable for an interface (implementations may require keys), but the interface package itself does not request or store secrets.
Persistence & Privilege
okThe skill does not request persistent/always-on presence (always: false) and does not modify other skills or system settings. Autonomous invocation is allowed by default (disable-model-invocation: false) but this is standard and not a red flag here.