CL1 Base

Public Codex skill for building and adapting CL1 human-neuron model workflows.

It focuses on:

• CL1-safe stimulation and near-limit validation
• CL1 transport and multi-port protocol patterns
• feedback design, including reward and surprise scaling
• readout and ablation guidance
• operational patterns such as recording, event signaling, caching, and observability

This package is designed for AI coding agents. The human-facing overview is here in README.md; the agent-facing instructions live in SKILL.md.

Contents

• SKILL.md: main Codex skill definition
• agents/openai.yaml: Codex UI metadata
• references/: CL1 protocol, modeling, and operations guidance
• scripts/: reusable helpers for safety validation, transport schemas, feedback scaling, caching, and training scaffolding

Install In Codex

Option 1: Manual install

Copy this folder into your Codex skills directory:

$CODEX_HOME/skills/cl1-base

If your environment uses the default Codex home layout, that is usually:

~/.codex/skills/cl1-base

Then restart Codex or start a new session.

Option 2: Install from GitHub with the skill installer

If your Codex environment includes the $skill-installer skill, install it from the public repo:

$skill-installer install <github-repo-or-path>

Examples:

$skill-installer install github.com/your-org/cl1-base-skill
$skill-installer install github.com/your-org/skills-repo/tree/main/cl1-base

After installation, start a new Codex session and invoke it with:

$cl1-base

Install In Claude Code

Claude Code does not use Codex SKILL.md folders directly. It uses Markdown subagents with YAML frontmatter.

Official Claude Code subagent docs:

• https://docs.anthropic.com/en/docs/claude-code/sub-agents

Project-level install

Create this file:

.claude/agents/cl1-base.md

User-level install

Create this file:

~/.claude/agents/cl1-base.md

Recommended conversion

Use this skill's SKILL.md as the system prompt body for the Claude subagent, and keep the same name and description.

Minimal Claude subagent template:

---
name: cl1-base
description: Build or adapt baseline CL1 human-neuron models, stimulation loops, and evaluation workflows. Use when Claude Code needs to design or modify CL1 training code, UDP stimulation and spike protocols, safe stimulation ranges, Same Frame Encoding (SFE), ablations, feedback-channel logic, or hardware-aware data and training loops for CL1 systems.
---

Paste the body of SKILL.md here, without the SKILL.md frontmatter.

If you want the Claude subagent to stay fully self-contained, fold the key material from:

• references/cl1-udp-protocol.md
• references/cl1-model-patterns.md
• references/cl1-operations-patterns.md

into the same Markdown file or keep those files nearby and reference them explicitly in the prompt.

Publishing Notes

• Keep the official Cortical Labs docs as the source of truth when values approach documented limits:
  • https://docs.corticallabs.com/
• Re-check the current docs before using near-limit amplitude, pulse width, frequency, or pulse-charge settings.
• If you publish this as its own repository, place the contents of this folder at the repo root.

Validation

Before publishing updates, run:

python path/to/quick_validate.py path/to/cl1-base
python scripts/cl1_transport_self_test.py
python scripts/cl1_udp_training_scaffold.py --self-test