I installed 901 Claude skills. My own audit tool graded my setup F.
retained text
I kept installing skills and MCP servers without ever counting what they cost. So I built a tool that measures it, and it graded my own machine F. The receipts: - 901 skills installed. Their descriptions alone add 39,402 tokens to every conversation, whether I invoke them or not. - 15 MCP servers probing for real: initialize + tools/list against each stdio server to get the true schema bill. - 92,413 tokens burned before my first message. - The same MCP server configured under 3 different names. A plaintext API key in an env block. A dead binary. The tool is ctxvitals, a vitals check for your agent stack. Audits skills, MCPs, rules, commands, and hooks across Claude Code, Cursor, Windsurf, Copilot, Codex, Devin. Flags secrets, exfil patterns, curl|sh, duplicates, broken configs. Gives you a grade and a shareable report. npx ctxvitals, zero deps. It also ships as a Claude skill, so your agent can audit itself. github.com/princejain756/ctxvitals Curious what your numbers look like.. I suspect 901 skills is not a record here.
document fetched 2026-10-13 00:41 · canonical https://reddit.com/r/ClaudeCode/comments/1x3zp3c/i_installed_901_claude_skills_my_own_audit_tool