HarnessKit and your files

HarnessKit is an editor for your agents’ own directories, not a host for them. Each agent keeps reading its own files, in its own format, whether HarnessKit is running or not. HarnessKit reads those files to show you what is installed. It writes to them when you install, enable, disable or delete something.

HarnessKit keeps one file of its own that matters: ~/.harnesskit/metadata.db. It holds facts that have no place in an agent’s files, such as where a skill came from and its audit results. It also holds snapshots, exact copies of MCP servers and hooks you disabled on agents that have no off switch of their own. No agent reads this file.

One test shows the model at work. Remove HarnessKit, and every agent keeps the setup it had. The exceptions are the items you had turned off, which Managing files in place lists.

Read in this order

Read the first four pages in order, since each builds on the one before it. The last three stand on their own.

Managing files in place, and what goes online What HarnessKit changes on disk, what it stores, and every reason it goes online.

Installing a skill: a directory copy What HarnessKit writes when you install a skill to another agent.

Installing an MCP server: per-agent format translation How one server entry is rewritten into each agent's config format.

Installing a hook: event translation How hook events are renamed between agents, and where that falls short.

How the security audit scores The audit rules, how points are deducted, and what the audit can't catch.

Marketplace: three sources and the CLI crawl Where Marketplace listings come from and how the CLI registry is curated.

Tip of the Day: an online file and a weekly crawl Where Tip of the Day comes from and how the tips are chosen.