agentleFS
BlockKnown publisherblock/buzz-app/src/buzz-mcp/AGENTS.md

buzz-app / buzz-mcp

AGENTS.md82 starsChanged 41 days agoApache-2.0View on GitHub

The tools in this directory are the reference interface for Buzz agents, local and remote. Remote agents run on another host with similar tools, so a prompt or skill that works for a local agent works for a remote one. Change them deliberately. - tools.ts: each tool's name, description and schema in TOOLS, and its behavior in HANDLERS. - attention.ts: the attention tools (interest*, watch*). - session-tools.ts: what a session lists, each tool's load (always or search), and the server…

Files
AGENTS.md

Buzz tools

The tools in this directory are the reference interface for Buzz agents, local and remote. Remote agents run on another host with similar tools, so a prompt or skill that works for a local agent works for a remote one. Change them deliberately.

Files

  • tools.ts: each tool's name, description and schema in TOOLS, and its behavior in HANDLERS.
  • attention.ts: the attention tools (interest_*, watch_*).
  • session-tools.ts: what a session lists, each tool's load (always or search), and the server instructions that cover the whole tool set.
  • prompt.md: the posting conventions every agent gets after the shared base prompt, src/features/agent-harness/base_prompt.md. Both name tools.

Interface

A tool's interface is its name, description, arguments (names, types, descriptions, which are required), defaults such as channel and reply defaulting to the current conversation, result shape, error text and load mode.

Design rules

  1. Many small tools are fine as long as each has a clear name, a precise description and explicit parameters. Prefer canvas_get and canvas_set to one canvas(action, …) tool.
  2. Don't hide tools behind a discovery meta tool, such as a search_tools or run_tool(name, args) wrapper. List each tool with its real schema, and use load: "search" for tools that don't need to load up front.
  3. Don't bloat descriptions or schemas. A description says what the tool does and anything the model can't guess. Routing, etiquette and workflow guidance go in the server instructions or prompt.md, not in every tool. Describe an argument only when its name and type don't make its meaning clear.
  4. Mark writes accurately with MCP tool annotations where possible.
  5. Use server instructions for guidance that covers the whole tool set: which tool to reach for, how tools relate, and shared conventions.

Remote agents

A remote agent's tools can differ: its host may add tools or arguments, for example for its sandbox, or leave out an argument it can't support. Keep the differences small so agents work the same way on any host.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.