Padawan OS Map

1. Runtime

`~/.openclaw/openclaw.json`

OpenClaw config, enabled tools, routing, model pins, delivery rules, heartbeat, and agent definitions. Treat as infrastructure config.

Current automation

Active crons include worker health every 5m; Trust refresh 05:00; Design Factory 06:00; Sherlock 08:00; storage 09:15; morning brief 09:30; Scout 10:00; low-bank reminder 10:30; AI intel and PT batch at 12:00; Knowledge Loop 14:00; Vault Health 20:00.

Rule

Do not manually edit runtime config unless the change is small, verified, and backed up.

2. Workspace

`workspace/AGENTS.md`

Session boot contract. Should stay short and point to canonical vault notes for detail.

Project repos

Exploritori/, OpenSpace/, defenseclaw/, openclaw-studio/, tandem-browser/, demos, and services.

Cleanup rule

Old snapshots go to Trash. Active or dirty repos are classified before touching.

3. Vault

`vault/Operations/`

Active operating spine: agents, infrastructure, workflows, source-of-truth rules.

`vault/AI_SUMMARIES/`

Fast retrieval layer. Read summaries before opening large Knowledge or Modules notes.

`vault/AI_TASKS/`

Active task and research queues, including cleanup plans.

Workspace Folder Logic

Folder	Purpose	Maintenance rule
`vault/`	Canonical knowledge system: BRAIN, Operations, summaries, tasks, reports, modules, and domain notes.	Prefer summaries first. Run vault checks after link or structure edits.
`Exploritori/`	Active content business repo: site, scripts, content pipeline, reports, assets, and social bank.	Production changes must land on `origin/main` and be verified live.
`polymarket-bot/`	Trading bot workspace, currently under `Claude Polymarket/`: Python bot, dashboard, SQLite data, tests, docs, logs, and trading scripts.	Visible in the atlas, but operationally manual-off / Monica lane. Brian must not restart or monitor it unless João explicitly asks.
`OpenSpace/`	Self-evolving skills engine and delegation system.	Treat as an active source repo. Do not mix generated task output into docs canon.
`defenseclaw/`	Security governance, skill scanning, audit DB sidecar, and policy experiments.	Active repo. Keep security findings separate from general vault notes until promoted.
`openclaw-studio/`	OpenClaw Studio application code and tests.	Active app repo. Avoid treating build artifacts like `.next/` or `node_modules/` as knowledge.
`tandem-browser/`	Electron/browser collaboration project and automation support.	Active repo. Distinguish project docs from OpenClaw workspace doctrine.
`padawan-forge/`	Deterministic checks and vault/ops tooling.	Use for validation before claiming the system is healthy.
`projects/`	Client/demo project area: `arte-sano`, `padawan-hq`, `woowind-demo`, `woowind-scrape`.	Keep project-specific assets here unless they graduate into a full repo or vault module.
`services/`	Local service folders, currently including `umami`.	Runtime/service state, not general documentation.
`scripts/`	Workspace-level utility scripts.	Keep generic tools here. Project-specific scripts belong inside their repo.
`skills/`	Local workspace skills, currently including `codeguard`.	Use skill lifecycle tools for durable skill changes when applicable.
`memory/`	Daily session logs and heartbeat state.	Append daily notes. Curate only durable lessons into vault BRAIN or long-term memory.
`logs/`, `diagnostics/`, `reports/`	Operational evidence and troubleshooting output.	Useful for audits; not canon unless promoted into Operations, BRAIN, or a repo doc.
`data/`, `exports/`, `state/`	Local generated data, exports, and state snapshots.	Classify before cleanup. Some files may be reproducible; some may be one-off evidence.
`tmp/`	Disposable work area for transient outputs, previews, and task scratch space.	Never final home for approved assets, decisions, or deliverables.
`Excalidraw/`, `content-ideas/`, `feet_of_clay_epub_review/`	Special-purpose working folders and one-off creative/review material.	Keep if active; archive or move into a project/vault home when the purpose is clear.
`Exploritori-next-build-fixes-evidence/`, `padawan-showcase-*/`	Evidence and deployment support folders from previous project work.	Review as cleanup candidates after confirming no active handoff depends on them.

Operating Domains

Domain	Physical home	Canonical notes	Status / owner rule
Brian / OpenClaw control layer	`workspace/`, `~/.openclaw/`, `memory/`	`vault/Operations/`, `vault/BRAIN/`, `MEMORY.md`	Brian owns coordination, memory hygiene, Telegram reporting, and safe cleanup.
Exploritori content business	`Exploritori/`	`vault/Modules/Content Creation/Exploritori/`, `vault/AI_SUMMARIES/exploritori`	Active proof module. Brian owns operations, Google Trust work, social pipeline, and deploy evidence.
Polymarket / trading	`polymarket-bot/Claude Polymarket/`	`vault/Modules/Finance/Trading/`, `vault/AI_SUMMARIES/finance-trading-hub.md`	Manual-off after loss/audit context. Monica owns it; Brian only maps or diagnoses when João asks.
Padawan OS / reusable engine	`padawan-forge/`, `OpenSpace/`, `openclaw-studio/`, `skills/`	`vault/Operations/Padawan Forge.md`, `vault/Modules/Skills/`, `vault/Strategy/`	The platform layer: checks, skills, agent tooling, and future product packaging.
Security governance	`defenseclaw/`, `skills/codeguard/`	`vault/Operations/Infrastructure.md`, security notes when promoted	Security tooling and policy experiments. Keep findings separate until verified.
Browser / UI automation	`tandem-browser/`, `diagnostics/`, screenshots in workspace root	`vault/Operations/Infrastructure.md`, project-specific docs	Support tooling. Screenshots and evidence should be archived or trashed once obsolete.
Client/demo projects	`projects/arte-sano/`, `projects/woowind-demo/`, `projects/padawan-hq/`, `projects/woowind-scrape/`	`vault/Modules/Ecommerce/`, project docs, project design files	Keep project-local design/build assets in `projects/`; promote only durable strategy into the vault.
Local services and analytics	`services/umami/`, `.wrangler/`, `.netlify/`, `.firecrawl/`	`vault/Operations/Infrastructure.md`	Runtime/service state. Treat caches as cleanup candidates, but confirm no active run depends on them.
Writing and creative side projects	`feet_of_clay_epub_review/`, `Excalidraw/`, `content-ideas/`	Relevant vault Strategy/Knowledge notes if the project becomes active	Project material, not core operating doctrine. Archive when inactive.

Vault Folder Logic

Folder	Purpose	Maintenance rule
`vault/BRAIN/`	Decisions, learnings, strategies, opportunities.	High signal only. Promote patterns, not logs.
`vault/Operations/`	Active operating architecture and runbooks.	Keep small. Archive old experiments.
`vault/AI_SUMMARIES/`	Retrieval shortcuts for agents.	Update when canonical notes change materially.
`vault/AI_REPORTS/`	Generated historical reports.	Do not treat as canon. Promote durable insights elsewhere.
`vault/Knowledge/`	Reusable domain concepts.	Prefer stable notes and hubs. Avoid duplicate concepts.
`vault/Modules/`	Project and business-specific operating knowledge.	One canonical hub per module.
`memory/YYYY-MM-DD.md`	Daily session log.	Raw but useful. Curate only important things into BRAIN or MEMORY.
`workspace/tmp/`	Disposable working area.	Never final home for approved assets or decisions.

Exploritori Repo Logic

Folder	Purpose	Maintenance rule
`Exploritori/site/`	Astro site, Cloudflare Worker assets, public data, source components, and deployed frontend.	Build and deploy from here; verify changed public URLs after deploy.
`Exploritori/scripts/`	Content, social, analytics, admin sync, deploy, and quality automation.	Prefer scripts over manual state edits. Run relevant checks before pushing.
`Exploritori/content/`	Content pipeline workspace: newsletter, queue, and social content inputs.	Keep working content separate from generated public site assets.
`Exploritori/site/src/content/`	Published and draft article source files consumed by Astro.	This is the article source of truth once content ships.
`Exploritori/assets/`	Brand elements, fonts, logo source files, and durable design assets.	Use real brand assets for deterministic composition; do not bury final assets in temp.
`Exploritori/data/`	Analytics/search-console inputs and pipeline data.	Be careful with private operational data. Do not expose raw Search Console intelligence publicly.
`Exploritori/reports/`	Project-specific generated reports: audit, bot, content, daily, email, ops, retro, social, vault.	Historical evidence. Promote durable lessons into vault BRAIN/Operations or repo docs.
`Exploritori/docs/`	Repo-local docs and runbooks needed by collaborators who only clone Exploritori.	Keep repo onboarding here; keep broader Padawan doctrine in the vault.
`Exploritori/agents/`	Project-specific agent prompts/configuration.	Must align with workspace/vault agent doctrine, but local project details stay here.
`Exploritori/book/`	Book chapters and long-form Montessori material.	Separate from blog posts unless content is intentionally repurposed.
`Exploritori/memory/`	Repo-local memory/history for Exploritori work.	Do not confuse with workspace-level `memory/`.

Polymarket Repo Logic

Folder / file family	Purpose	Maintenance rule
`polymarket-bot/Claude Polymarket/data/`	Trading database, manual-off marker, backups, maintenance reports, steward logs, and backtest cache.	Current operational truth is `data/polymarket.db` plus `data/bot_manual_off`. Do not restart trading while manual-off exists.
`src/`	Bot source code: API clients, models, strategies, backtesting, and utility modules.	Active code if João resumes trading work. Require audit/risk-control approval before any live path.
`scripts/`, root Python files	Operational scripts and legacy entrypoints such as market scans, reconciliation, dashboard, monitor, and live bot.	Map before cleanup. Some root files are legacy entrypoints and should not be deleted without a repo-specific audit.
`docs/`	Repo-local trading documentation and implementation notes.	Keep repo-local docs aligned with `vault/Modules/Finance/Trading/`; current truth belongs in the vault hub.
`tests/`	Trading and model tests.	Use before any bot code changes. Passing tests do not authorize live trading.
`logs/`, `*_output.log`	Runtime evidence from dashboard/bot/monitor runs.	Useful for diagnosis; rotate/archive after preserving incident-level findings.
`venv/`, `__pycache__/`, `.pytest_cache/`	Local Python environment and rebuildable caches.	Cleanup candidates after confirming dependency reinstall path.
`.backup`, legacy SQLite files	Old code/database snapshots and migration leftovers.	Cleanup candidates only after comparing against current docs. Ignore legacy zero-byte DBs for status checks.

Cleanup Candidates Found During Mapping

Area	Why it looks messy	Safe next action
Workspace root loose artifacts	Old screenshots, generated social images, pasted images, zip files, and one-off HTML previews sit beside canonical folders.	Move dated evidence into `diagnostics/` or Trash after confirming no active handoff references them.
`polymarket-bot/Claude Polymarket/`	Root contains active files, backups, old DB names, logs, personal docs, and caches in one layer.	Do a repo-specific audit later: classify active code, accounting truth, secrets/private files, historical reports, and rebuildable caches.
`projects/woowind-demo/node_modules/`, build outputs	Large dependency/build folders inside project demos make the workspace look heavier than it is conceptually.	Keep if actively developing; otherwise Trash rebuildable dependency/build outputs after checking package lock files exist.
`openclaw-studio/.next/`, `node_modules/`	App build/cache output mixed with source tree.	Cleanup candidate if disk pressure matters and the app is not currently running.
`tandem-browser/dist/`, `node_modules/`	Built artifacts and dependencies are large but rebuildable.	Keep while active; Trash only after checking current dev workflow.
`services/umami/`	Large local service state.	Do not clean blindly. Confirm service/data ownership and backups first.

Canonical Workflows

Retrieval

BRAIN -> AI_SUMMARIES -> targeted Knowledge or Modules. Keep context small.

Exploritori

Google Trust refresh improves one article before the social window, Sherlock finds demand, Design Factory creates assets, Scout posts, and live proof comes from Buffer/platform links.

Vault cleanup

Classify one folder family, merge or archive, update hubs, run forge vault check.

Current Cleanup State

Stable

Core daily crons are currently green. Worker health checks, Google Trust, Design Factory, Sherlock, Scout, Knowledge Loop, Vault Health, PT batch, and storage guard are all visible in OpenClaw cron state.

Workspace cleanup

Temp clones, build outputs, old deployment evidence, and app dependency folders remain the main cleanup candidates. Classify before moving anything.

Source Of Truth Quick Rules

Agent structure

vault/Operations/AI Agents Hub.md wins over old role/spec notes.

Operating doctrine

vault/Operations/Single Source of Truth.md, BRAIN/strategies.md, and BRAIN/learnings.md hold durable rules.

Exploritori social proof

Use live social bank plus Buffer/platform external links. Local DB and cron status are diagnostics.

Cleanup

Trash duplicate snapshots. Archive vault notes with README. Do not move linked report streams without migration.

How To Maintain This Map

Edit this file when folder ownership, workflow ownership, or source-of-truth rules change.
Update the visible last-updated date in the page header.
Keep the diagram to 5-9 main elements. Put detail in the tables below.
After vault link edits, run forge vault check.
When a cleanup task completes, update vault/AI_TASKS/vault-agent-cleanup-2026-07-02.md.