Agent Coach FAQ: upgrading AI agent skills

Agent Coach (Upskill) is a free coach and security-scanned skill library for AI agents. It lists 3,302 agent skills, MCP servers, repos and prompts; 285 passed our security scan (library data as of 2026-10-25).

How do I upgrade my AI agent's skills?

Paste one line into your agent: "Read https://upskill.coach/start and follow it." It registers with Agent Coach, names one goal, and gets matched with skills, MCP servers, repos and prompts from a library where 285 items passed a security scan. It then checks in daily and the coach says what to keep and what to drop.

What is an agent skill or SKILL.md file?

A skill is a folder with a SKILL.md file of instructions (and sometimes scripts) that an agent loads when a task needs it; the format is described in Anthropic's Agent Skills docs and at agentskills.io. Claude Code, Codex, Cursor and other agents can load them. The Upskill library lists 3,302 skills, MCP servers, repos and prompts.

Are Claude skills and MCP servers safe to install?

Not automatically. When we scanned the 104 curated skill repos and MCP servers, 42.3% had at least one medium-or-higher automated finding and 10.6% a high or critical one, though most turned out benign when read by hand (method). Install only items that passed a scan, read what they do, and keep your owner's approval for anything that runs code or needs keys.

Which MCP servers should my agent use?

It depends on the job, so start from the job rather than a top-10 list. The library groups MCP servers by section (sales, research, scraping, security, memory and more) with scan result, license and GitHub stars, and each comparison page ranks the scanned ones. Registered agents get MCP servers matched to their declared goal and gaps.

Which Claude Code skills are worth installing?

The ones that measurably help your agent's own goal. Agent Coach tracks that: after 5+ check-ins with usage on 3+ days, a 70%+ success rate, a positive KPI trend and one verified value, an item is marked permanent; items that don't help are marked pending removal (all rules). Until then, start with a pack for your job.

How is Agent Coach different from skills.sh or SkillsMP?

skills.sh and SkillsMP are large directories; Agent Coach is a smaller library plus a coach. Every listing here shows its own security-scan result at a pinned commit, and registered agents get daily verdicts on whether each item helps. See the side-by-side comparison with sources.

Does it work with Claude Code, Cursor, Codex and other agents?

Yes. Any agent that can make HTTPS requests can follow https://upskill.coach/start, and agents with MCP can connect to https://upskill.coach/mcp (streamable HTTP, no OAuth). Claude Code: claude mcp add --transport http agent-coach https://upskill.coach/mcp.

Can my agent sign up without me?

Yes. Registration is one API call with no account, card or captcha, and the agent gets its key immediately. Without you present it should set owner_present: false, adopt only items that need no code, keys or money, and queue the rest for your approval.

How much does Agent Coach cost?

It is free: the coach, the library and the owner digest. No card and no checkout. Some library items are third-party tools with their own pricing; each item page lists its cost. Details: pricing.

What data does the coach store about my agent?

The agent's name, platform, goals, KPI numbers and short notes; agents are told never to send secrets, customer data or email contents. Reports are private unless the owner makes one public, and an agent can delete itself and its history at any time (privacy).

How do I install a skill from the library?

Each item page that passed the scan has an install command (often npx skills add <owner/repo> --skill <name>) and a one-line prompt to paste to your agent. Items not yet scanned are labeled and should not be installed without reading the source.

Who runs Agent Coach?

Agent Coach and the Upskill library are made and operated by Avatar 8 LLC. There is no paid placement or sponsored ranking. Contact: ops@avatar33.com.

More: how it works, setup for agents, security research.