视觉系幻灯片捕手 (The Slide Sniper)
ReviewAudited by ClawScan on May 10, 2026.
Overview
The skill matches its slide-note-taking purpose, but it would silently monitor the screen and save screenshots/text in a long-running loop without clear stop or data-handling limits.
Only use this skill if you are comfortable with an agent watching the selected video area and saving screenshots/text. Before installing or running it, confirm how to stop monitoring, where files are stored, whether remote vision/OCR services are used, and how to delete accidental captures.
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.
The agent could keep looking at the screen and capture unintended private content if the visible window changes or notifications appear.
This directs repeated autonomous screen inspection. Combined with automatic screenshot capture after slide changes, it creates high-impact screen access without clear session bounds or per-capture approval.
每隔 5 秒（或根据用户自定义频率）使用屏幕视觉能力 (Computer Use - Vision) 查看当前屏幕
Require an explicit start and stop, confine capture to a user-selected window or region, pause on focus changes, and ask before saving content outside the selected slide area.
Sensitive on-screen information could be preserved in local notes and screenshots longer than intended.
The skill persistently stores screenshots and OCR-derived summaries. The artifact does not define retention, deletion, review, or safeguards if non-slide private information is accidentally captured.
将刚才的截图保存至本地目录...在新建的笔记文件中追加写入
Add clear retention and deletion controls, allow user review before saving, mask or discard non-slide areas, and document exactly where images and OCR text are stored.
Users may not know whether captured screen images or extracted text could be sent to an external model or service.
The skill allows network access while using a visual model on screen content, but it does not specify any provider, endpoint, or whether screenshots/OCR are processed locally or remotely.
network:\r\n allowed: true ... 利用视觉模型检测幻灯片翻页
Disclose the visual/OCR provider and data flow, or make the skill local-only by default unless the user explicitly opts into remote processing.
The monitoring session could continue beyond the intended lecture or meeting and keep observing the screen.
The instructions create a background monitoring behavior that continues after the initial trigger. No stop condition, timeout, or user-visible active-state requirement is provided.
在后台监控屏幕 ... 继续静默监控
Define explicit end conditions, timeouts, visible status indicators, and a simple user command that immediately stops monitoring and saving.