KOGNIT1V Cognitive Skills Pack

Neurodivergent tools for neurodivergent people.

Portable Codex skills for nonlinear builders working with AI.

KOGNIT1V packages a set of agent skills for people who want AI collaborators to handle branching thought, fast context shifts, long memory, taste, tempo, and proof. It is designed for local-first agent workflows: install the skills, open a project, and give the agent a better operating posture before it starts changing files.

This repository contains Markdown-based Codex skill instructions plus validation scripts. Install them locally to guide agent behavior in your own projects.

What It Gives An Agent

• Guides agents through unfamiliar repositories before broad file search.
• Provides reusable thinking patterns instead of one-off answers.
• Gives parent and worker agents clearer delegation contracts.
• Preserves structured handoffs across long sessions.
• Prompts reputation checks around commits, READMEs, and releases.
• Helps preserve human voice when turning rough thoughts into usable text.
• Turns mental overflow into reusable external structure.

Use It When

• You want an agent to review or modify an unfamiliar repo without starting blind.
• You need to turn a rough idea into a reusable generator, brief, or operating pattern.
• You are preparing public repo copy, a PR, or a release and want less hype and more trust.
• You need a long-running agent session to leave a handoff that another agent can reload.
• Your mind moves faster than your tools can preserve.

Compatibility

Surface	Status
Codex local skills	Supported
Codex plugin manifest	Included
Node.js	>=20
m1nd	Optional, checked by bootstrap
Secrets	Env names only, no values
Install proof	npm run smoke:install
Toolchain boundary	docs/TOOLCHAIN.md

Included Skills

Skill	Helps with	Use when
kognit1v	Routing intent through the cognitive stack	A request needs patterns, proof, handoffs, delegation, and voice together
pattern-architect	Turning ideas into reusable generators	You need briefs, taste profiles, prompts, PRDs, or anti-pattern maps
proof-grown-systems	Separating claims from evidence	A prototype, feature, or method needs contracts, proof, and non-claims
tempofastlane	Delegation and temporal calibration	Parent/worker agents need bounded tasks and realistic execution loops
fastlane	Fast bounded implementation	A worker lane should make a scoped code change
m1nd-first	Graph-first repo investigation	An agent should inspect structure before noisy search
m1nd-operator	m1nd operation and recovery	m1nd is available and the agent needs trust, ingest, routing, or graph diagnosis
pathos	Continuity handoffs	A long session needs transferable state and relationship context
dext3r	Local companion memory and perspectives	A Codex chat should bind to the local DEXT3R companion, session memory, or Council-style perspectives
brotherizer-codex-runtime	Human voice preservation	Rough text needs to become direct, natural, technically precise communication
git-postman	Public repo reputation	Commits, PRs, READMEs, release notes, or pushes need credibility checks
omx-codex-app-bridge	Local Codex bridge workflow	A session needs to connect safely with an existing local Codex bridge
universal-triple-flow	Generic multi-agent flow	A director thread coordinates several bounded worker lanes

This repo does not include private product systems, unreleased project doctrine, credentials, or generated project artifacts.

What Actually Installs

KOGNIT1V installs the cognitive layer: Codex skills, contracts, examples, validation scripts, and agent operating docs. It does not install Codex itself, API keys, private projects, hosted providers, or every local tool Max uses.

Optional tools such as m1nd, DEXT3R, tmux, Cargo/Rust, Python helper runtimes, and local voice/bridge workflows are treated as host capabilities. The bootstrap can check them and, for m1nd only, prepare installation when explicitly asked.

npm run doctor:tools

If a tool is missing, the agent should name the capability gap and continue with the safest available path. No fake claims.

Install

Check the package:

npm run validate
npm run smoke:install

Install into Codex:

npm run bootstrap -- --install-skills --codex-home "$CODEX_HOME"

Or copy manually:

cp -R skills/* "$CODEX_HOME/skills/"

If you use m1nd, KOGNIT1V can check or prepare it explicitly:

npm run bootstrap -- --with-m1nd --install-m1nd

Agent Entry

Agents should start with:

1. AGENTS.md
2. skills/kognit1v/SKILL.md
3. npm run validate

The repo is intentionally shaped for agents first. Humans can read it, but its main job is to give an AI collaborator reliable operating instructions.

Scope

KOGNIT1V is a public cognitive skills pack. It is not a full app generator, a hosted service, or a bundle of Max's private projects.