Assistant CLAUDE.md

Repository instructions for Mewbo, an AI assistant built as a conversation state machine with a system for managing sub-agents. They also explain its monorepo structure, where multiple related projects share one repository.

In plain words
What is it for?
Hydrating project knowledge before making changes, understanding Mewbo’s agent system, and adding modules without breaking its monorepo layering rules.
Why use it?
They tell an agent which project knowledge to load first and how to keep code dependencies within the required architecture.

Instructions file

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/bearlike/assistant/claude-md
Clone the repo
git clone --depth 1 https://github.com/bearlike/Assistant
Per session 6,737 This file is loaded in full into every session.
When invoked 6,737 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.06737 $0.06737
Opus 5 $0.03368 $0.03368
Sonnet 5 $0.01347 $0.01347
Haiku 4.5 $0.00674 $0.00674

Measured 3d ago against content hash e01a86d81b97, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

Assistant CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 229 lines

How it starts

The opening of the file, as written. The whole thing — 229 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agents Guide — Mewbo

MANDATORY: Hydrate before touching files

At the start of every conversation and every non-trivial task, ask a configured code-wiki MCP tool a question about bearlike/Assistant — try the primary configured one first, falling back to a secondary if one is configured. Use read_wiki_structureread_wiki_contents for deeper exploration. Only read local files after hydration. Include this directive in subagent prompts.

Wikis lag the repo. Use them for intuition; verify structure against local source.

What Mewbo is

An AI assistant modelled as a conversation state machine with a hierarchical agent hypervisor. Core engine: a single async ToolUseLoop bound to an LLM via native bind_tools. Sub-agents admitted via spawn_agent, tracked by AgentHypervisor, resolved into one of four terminal states (completed, failed, cancelled, rejected). Interfaces: CLI, web console, REST API, Home Assistant, Nextcloud Talk — all share the core engine. For architecture details, query the wiki.

Monorepo layering — read before adding a module

Dependencies flow strictly down this DAG; never up, not even a lazy try/except ImportError:

mewbo_core (lean SDK) → mewbo_tools · mewbo_graph · mewbo_iam → apps
  • mewbo_core — lean orchestration SDK: generic, dependency-light primitives only. No product/graph code; heavy optional deps go behind a mewbo-core[...] extra.
  • mewbo_tools — subprocess/remote integrations (MCP, LSP, file edit). Deps core only.
  • mewbo_graph — optional capability library (graph, memory, embedding, SCG). Heavy deps behind extras + import-guards. Deps core only; never an app.
  • mewbo_iam — optional identity kernel (principals, authenticators, roles, teams, grants, audit). Base deps are core + pydantic and NOTHING else; the network/crypto provider legs sit behind the oidc/ldap/saml extras. Deps core only; never an app.
  • apps — thin product surfaces: HTTP routes, wire contracts, transport, glue. Compose libraries; never host a reusable engine.

Read the full file on GitHub · 229 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago First seen · 229 lines · 6,737 tokens per session scan A e01a86d81b97

Subscribe to this mod's changes

Assistant CLAUDE.md is an instructions file published in the GitHub repository bearlike/Assistant (41 stars, last pushed 9d ago), licensed MIT. It adds 6,737 tokens to every session, about $0.0337 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.