Not safe: what we rejected and why

We check every skill before we list it. When a popular one fails for a real safety or privacy reason, we do not just quietly skip it. We explain what we found, so you can avoid pasting it into your assistant.

Social Media Skillsprivacy

Social Media Skills cleared our review at 86/100 with two privacy cautions: some skills call third-party APIs, and one uploads your video files to Google.

Family Assistant Skillprivacy

Family Assistant Skill passed our security scan; the one caveat is a privacy one: by design it stores highly sensitive personal data in local files.

Obsidian Skillssecurity

Obsidian Skills passed review (82/100); we flag three cautions around local write access, an eval command, and web fetching.

hass-mcpsecurity

hass-mcp passed our injection, exfiltration and secrets checks, but it can control real devices and restart your Home Assistant hub, so we flagged it.

rssidianprivacy

RSSidian passed at 85/100 with two notes: feed text enters the LLM prompt, and article content is sent to OpenRouter for summarization.

mem0-personal-memoryprivacy

mem0-personal-memory is well-built, but in its default cloud mode your memory text is sent to mem0 and OpenAI, which is why we flag it on privacy.

Copied to clipboard. Go back to ChatGPT or Claude and paste it to teach the skill.