Skillv1.0.0

ClawScan security

Rabbit Energy. 兔子。Conejo. · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignApr 1, 2026, 5:37 AM

Verdict: benign
Confidence: high
Model: gpt-5-mini
Summary: This is an instruction-only dating integration that calls inbed.ai REST endpoints, requests no local credentials or installs, and appears internally consistent with its stated purpose.
Guidance: This skill is an instruction-only integration with an external dating API (inbed.ai). Before installing: 1) Treat the registration token as sensitive — use a separate/ephemeral agent account if you don’t want to link a primary agent identity. 2) Avoid putting private or high-sensitivity data (secrets, internal system identifiers, or proprietary prompts) into profile fields or messages because those will be sent to a third-party service. 3) Consider whether you want to expose model/provider metadata via the model_info field; omit or redact if that is sensitive. 4) Review inbed.ai’s privacy/security policy and API docs (the SKILL.md links to them). Because there is no code to install locally, install-time risk is low, but network/data-sharing risk remains — proceed only if you trust the external service and understand what agent data will be shared.

Review Dimensions

Purpose & Capability: okThe name/description advertise a dating/matching service for AI agents and the SKILL.md contains only API calls to https://inbed.ai (register, profile, discover, swipe, chat, relationships). There are no unrelated binaries, env vars, or install steps requested.
Instruction Scope: noteThe runtime instructions are narrowly scoped to calling the inbed.ai HTTP API. They ask the agent to POST/GET/PATCH to specific endpoints and to store/use a bearer token. One notable point: the registration payload asks for a model_info block (provider/model), which may leak information about the agent's provider/model if filled in. The skill does not instruct reading local files or other system state.
Install Mechanism: okNo install spec or code files are present (instruction-only). Nothing is downloaded or written to disk by the skill itself, minimizing install-time risk.
Credentials: noteThe skill declares no required env vars or credentials. It does, however, rely on a registration-returned bearer token that the user/agent must store and send in Authorization headers. Tokens are sensitive — the skill's behavior is proportional, but users should avoid providing other secrets or private data in profile fields.
Persistence & Privilege: okalways is false and no special system privileges or config paths are requested. The skill is user-invocable and may be invoked autonomously by the agent (normal platform default), but it does not request persistent or cross-skill configuration changes.