# CE System Export
*Exported: 2026-03-13 | Source: /root/.openclaw/workspace*

This is a structured snapshot of the Curious Endeavor AI agent system.
It contains everything needed to understand, document, and recreate the system.

**Nothing was deleted from the source. This is a copy-only export.**

---

## What's In Here

```
system-export/
├── README.md                        ← This file
│
├── agents/                          ← Identity files for every agent
│   ├── kitt/                        ← Orchestrator / CEO
│   ├── anton/                       ← Critic / Quality gate
│   ├── gerri/                       ← PM / Pipeline coordinator
│   ├── jessica/                     ← Art direction / Layout
│   ├── julia/                       ← Research / Publishing
│   ├── kittoro/                     ← eToro Discord server agent
│   ├── ogilvy/                      ← Copywriter
│   ├── tatiana/                     ← Visual / Image generation
│   ├── thibault/                    ← Dev / Engineering
│   ├── unagi/                       ← (role TBD)
│   └── _deprecated/                 ← Legacy agent stubs (not active)
│
├── diagrams/
│   └── agent-interaction-diagram.md ← ASCII map of how agents interact
│
├── config/                          ← System config and rules
│   ├── skills-registry.md           ← Master skill list + status
│   ├── learnings.md                 ← Rules accumulated from mistakes
│   ├── reference.md                 ← Project refs (API keys redacted)
│   ├── WORKFLOWS.md                 ← Cross-agent workflow docs
│   └── project-structure.md        ← Workspace layout explained
│
├── scripts/                         ← Active scripts only (referenced in HEARTBEAT/MEMORY)
│   ├── core/                        ← Heartbeat, email check, QMD, git backup
│   ├── discord/                     ← team-show.cjs
│   ├── figma/                       ← figma-relay, figma-relay-web
│   └── images/                      ← compress-images.js
│
├── skills/                          ← Agent skills, split by status
│   ├── active/      (7)             ← 🟢 Confirmed used in real work
│   ├── required/    (17)            ← 🟡 Required by system, not yet enforced
│   └── installed/   (33)            ← 🆕/🔴 Installed, not yet activated
│
├── styleguide/                      ← CE visual brand standards
│   ├── brand-framework.md
│   └── ce-ui-quality/
│
└── methodology/                     ← CE strategic methodology docs
    ├── ARCHITECTURE.md
    ├── architecture/
    ├── creative-pipeline/
    └── methodology/
```

---

## How to Recreate the System

### 1. Agent Setup
Each agent folder contains the files that define its identity and behaviour:
- `IDENTITY.md` — who the agent is, capabilities
- `SOUL.md` — personality, voice, operating rules
- `AGENTS.md` — boot sequence, task delegation rules
- `HEARTBEAT.md` — operational rules, what to do on each tick
- `TOOLS.md` — which tools the agent can use
- `USER.md` — profile of Assaf (shared across all agents)

Load these as system prompt context when spawning each agent.

### 2. Skill Loading
Skills live in `skills/`. The system uses them as on-demand instruction files:
- Agent gets a task → checks `config/skills-registry.md` for a matching skill
- Reads the matching `SKILL.md` → follows its instructions
- Start with `skills/active/` — these are proven working
- `skills/required/` are the methodology backbone — load them for their respective task types

### 3. Scripts
Core infrastructure scripts go in `/scripts/core/`. They're called by:
- Cron: `heartbeat-day.sh`, `heartbeat-night.sh`
- Kitt on session end: `qmd-compact.py`
- Kitt every heartbeat: `auto-compact-sessions.py`, `check-kitt-email.py`

### 4. Memory Architecture
| File | Purpose |
|------|---------|
| `MEMORY.md` | Critical live facts (lean, always loaded) |
| `LEARNINGS.md` | Rules from mistakes (loaded on boot) |
| `memory/qmd/current.json` | Active task working memory |
| `memory/YYYY-MM-DD.md` | Daily session logs |

---

## Key Design Decisions
- Kitt never touches creative work — all creative routes through Gerri → team
- Anton reviews inline (during work, not just at end)
- Sub-agent briefs are max 5 lines: What + Constraints + Autonomy + Output + Done-when
- All new Discord channels are private by default
- API keys and secrets are never in exported files — stored in `/home/clawd/secrets/`

---

## What's NOT in This Export
- `memory/` daily logs — session-specific
- `work/pitches/`, `work/active-projects/` — client work
- `google-auth/` tokens — live secrets
- `public/` — deployed web repo (separate git)
- API keys — redacted in config files