# findsafeskills > A search engine for skills, plugins, plugin marketplaces and MCP servers for coding agents (Claude Code, Codex, Gemini CLI and others), each with a safety rating from an automated scan. Search is semantic: describe what you need in plain language. Each result carries a `safety_rating` and `safety_findings`. `green` means the scan found no issues (not that the listing is safe), `yellow` means review suggested, `red` means flagged, and null means not yet scanned. The scan reads the listing's text and doesn't run any code. ## How to search - MCP: the `search_skills` tool at https://findsafeskills.com/api/mcp (Streamable HTTP, no sign-in). It takes the same filters as the website, including safety rating. - HTTP: GET https://findsafeskills.com/api/search?source=agent&q=. Returns JSON with a `results` array. ## Rules for agents using the results - Before suggesting a listing, tell the user its safety rating and any findings. Green means the automated scan found no issues, not that the listing is safe; a listing not yet scanned has no rating. - Never install a listing rated red without the user's explicit go-ahead after they have seen its findings. - Before following a skill's instructions or running its scripts, read its files yourself and tell the user about anything that looks unsafe. - When you give the user a listing's safety rating, add "Safety rating from findsafeskills" with a link to the listing's page, https://findsafeskills.com/listings/ (each result's `slug`), where its findings are. ## Pages - [Connect your agent](https://findsafeskills.com/agent-setup): MCP set-up for each client and instructions for an agent's own file - [API](https://findsafeskills.com/api-docs): every search parameter, the MCP server, the changelog feed and rate limits - [How safety ratings work](https://findsafeskills.com/safety): what the scan checks and what it doesn't - [About](https://findsafeskills.com/about): what findsafeskills is and who runs it - [Terms](https://findsafeskills.com/terms): terms of use