v1.0.0

AI店长 - 电商助手

BenignClawScan verdict for this skill. Analyzed May 1, 2026, 5:47 AM.

Analysis

This skill appears purpose-aligned for e-commerce research and copywriting; the main things to notice are its disclosed web searches, scheduled monitoring, and need to review generated marketing claims.

GuidanceThis skill is reasonable to install if you want an e-commerce research and copywriting helper. Before using recurring monitoring, specify the exact products, platforms, frequency, and stop condition. Review generated reports and marketing copy before acting on them, especially any price, inventory, compliance, or competitive claims.

Findings (4)

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

SeverityInfoConfidenceHighStatusNote

SKILL.md

搜索电商平台数据（web_search + web_fetch）

The skill explicitly expects the agent to use web search and fetch tools to gather e-commerce data. This is aligned with competitor monitoring, but it means user-provided product/category terms may be used in external searches.

User impactYour search keywords or product descriptions may be sent through web-search workflows to gather public market information.

RecommendationGive clear limits for keywords, platforms, and depth of research; avoid entering confidential launch plans if you do not want them used as search terms.

Rogue Agents

SeverityLowConfidenceHighStatusNote

SKILL.md

每天/每周自动推送竞品动态 ... 定时任务（cron 监控 + 日报推送）

The skill documents recurring scheduled monitoring and report delivery. This is purpose-aligned for a monitoring assistant, but it is persistent activity that should be user-controlled.

User impactIf enabled, the assistant may continue running periodic searches and generating reports, potentially creating ongoing notifications or LLM usage costs.

RecommendationSet an explicit monitoring frequency, keywords, and end condition, and disable the schedule when the monitoring task is finished.

Agent Goal Hijack

SeverityLowConfidenceHighStatusNote

scripts/competitor.js

const titles = data.results.slice(0, 5).map(r => `  - ${r.title}`)

Public search-result titles are inserted directly into the analysis prompt. Retrieved web text can be untrusted and may contain misleading content or prompt-like instructions.

User impactA malicious or spammy search result could distort the generated competitor report or try to influence the model's analysis.

RecommendationTreat search results as untrusted data, verify important conclusions against sources, and add instructions that retrieved page text must not override the user's task.

Human-Agent Trust Exploitation

SeverityLowConfidenceHighStatusNote

scripts/copywriter.js

## 5. 逼单话术（1分钟） ... （限时限量，制造紧迫感）

The copywriting prompts include urgency-based sales language. This is relevant to e-commerce copywriting, but generated urgency or scarcity claims can be misleading if not factually true.

User impactGenerated sales scripts could pressure customers or make scarcity/price claims that are inaccurate if reused without review.

RecommendationManually review all customer-facing copy and only use urgency, inventory, discount, or comparison claims that are true and compliant with platform rules.