v1.0.1

AI Meeting Room

BenignClawScan verdict for this skill. Analyzed May 1, 2026, 5:34 AM.

Analysis

This instruction-only skill is coherent and purpose-aligned, but users should notice that it automatically performs web searches and may reuse prior meeting context.

GuidanceThis skill appears safe to install as an instruction-only advisory tool. Before using it, remember that it may run web searches automatically and may continue from earlier meeting context; avoid sharing confidential information unless you are comfortable with that use, and verify important legal, financial, or strategic claims independently.

Findings (3)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

Abnormal behavior control

Checks for instructions or behavior that redirect the agent, misuse tools, execute unexpected code, cascade across systems, exploit user trust, or continue outside the intended task.

Tool Misuse and Exploitation

SeverityLowConfidenceHighStatusNote

SKILL.md

주제가 들어오면 회의 시작 전 **web_search 실행** ... **스프린트:** 3~5회 검색 ... **풀/딥:** 8~12회 검색 ... 사용자에게 안내 후 바로 시작 (확인 불필요)

The skill directs the agent to run multiple web searches automatically and begin without a separate confirmation step. This is purpose-aligned for data-backed meetings, but it is still external tool use users should notice.

User impactThe user’s topic may be sent as web search queries, and the agent may spend tool/cost budget on several searches.

RecommendationDo not include confidential business plans, private customer data, or sensitive internal details unless you are comfortable with them being used for web research.

Human-Agent Trust Exploitation

SeverityInfoConfidenceMediumStatusNote

references/agent-roles.md

각 에이전트는 **구체적 배경**을 가진다 ... **신뢰할 수 있는 전문가 페르소나**

The skill uses authoritative expert personas with prestigious backgrounds. The artifacts frame them as AI personas, so this is not deceptive by itself, but users should not treat the output as advice from real licensed professionals.

User impactThe simulated experts may sound more authoritative than the model’s actual reliability supports.

RecommendationUse the output as brainstorming or decision support, and verify legal, financial, medical, or investment claims with qualified professionals and primary sources.

Sensitive data protection

Checks for exposed credentials, poisoned memory or context, unclear communication boundaries, or sensitive data that could leave the user's control.

Memory and Context Poisoning

SeverityLowConfidenceMediumStatusNote

SKILL.md

연속 회의 체크: 이전 회의 맥락이 있으면 자동으로 이어받기

The skill tells the agent to reuse previous meeting context automatically. That can help continuity, but it can also carry forward stale, sensitive, or user-unreviewed context.

User impactInformation from earlier meetings may influence later advice, including details the user did not intend to reuse.

RecommendationAsk the agent to summarize the carried-over context before relying on it, and start a fresh meeting for unrelated or sensitive topics.