madcatone

Agentic Mode — Communication skill for Claude Code

Communication community

A documentation contract that makes any repo legible to any collaborator — human or agent, no chat history: layered ID-joined docs, a stdlib checker, a CI gate, plus 7 installable agent skills.

How to install Agentic Mode

This entry records only its repository, not the path inside it, so there is no exact command to give. Open madcatone/agentic-mode and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agentic Mode does

A documentation contract that makes any repo legible to any collaborator — human or agent, no chat history: layered ID-joined docs, a stdlib checker, a CI gate, plus 7 installable agent skills.

Alternatives in Communication

  • HolaOS — Open-source agentic workspace enterprises can make their own 10.9k ★
  • Chat Langchain — A documentation assistant deployed as a Managed Deep Agent 6.4k ★
  • Social Media Agent — 📲 An agent for sourcing, curating, and scheduling social media posts with human-in-the-loop 2.8k ★

README

agentic-mode

A **documentation contract** that makes any repository legible to any collaborator — human or agent, in any tool, with **no chat history**. It is a small set of layered, cross-linked docs joined by stable IDs, plus a config-driven checker and a CI gate that keep those docs from drifting out of sync with the code.

This repo is the **canonical source** of the methodology. Downstream copies (a Claude Code plugin, a Devin IDE rule, a vendored subtree in your own repo) should sync *from here* — see [Adapters](#adapters) and [Canonical source](#canonical-source).

Why

Chat history is not a durable substrate. The moment a new collaborator arrives — a different person, a different agent, a fresh session — everything that lived only in the conversation is gone. Teams paper over this with a `NOTES.md` that mixes contract, guide, status, and process until no one knows which section is authoritative, and it rots.

agentic-mode fixes that with three principles (full doctrine in [`doctrine/BOOTSTRAP-CORE.md`](doctrine/BOOTSTRAP-CORE.md)):

  1. Layered, non-overlapping docs. Each file has exactly one job; facts have a single canonical home and are cross-linked, not duplicated into independent authorities.
  2. Stable IDs for every behavior. Requirements carry PREFIX-001 IDs; the user guide, validation blocks, commits, and MR/PR notes all cite them. The ID is the join key that lets any reader jump across layers without chat history.
  3. Append-only provenance + per-feature done evidence. Iteration history records why a change landed (never rewritten); validation blocks record that it was actually done and how to re-verify.

The contract owns **What** (repo facts, behavior contract) and **Done** (acceptance evidence). It never dictates **How/Who** (which tool, model, or dispatch you use) — that stays with whoever does the work, which is what keeps the contract portable across every harness.

Quick start

You need only Pytho