os-guidelisted

# Agentic OS Guide The core insight: LLMs are stateless functions. `CLAUDE.md` is the only file loaded by default into every conversation. The **Agentic OS** pattern turns this constraint into a full operating system metaphor (kernel = CLAUDE.md hierarchy, RAM = context/ folder, disk = dated session logs, stdlib = skills/, processes = sub-agents, shell = slash commands, cron = /loop, boot = START_HERE.md, autoresearch loop = os-eval-runner). Full OS-concept mapping and the skill-category table (orchestration/evaluation/mutation/memory/reporting/bootstrap/ utility/diagnostic) are in `references/detailed-reference.md`. Dependencies (Python 3.8+, stdlib only) are also there. ## Execution Flow Execute these phases in order. Do not skip phases. This skill uses **Progressive Disclosure**. Load only what you need: 1. For CLAUDE.md scope rules and precedence -> read `references/architecture/claude-md-hierarchy.md` 2. For context/ folder patterns (soul.md, user.md, memory.md) -> read `references/architecture/context-folder-patterns.md` 3. For /loop and heartbeat.md scheduling -> read `references/operations/loop-scheduler.md` 4. For sub-agents, hooks, auto-memory -> read `references/architecture/sub-agents-and-hooks.md` 5. For memory hygiene (write/promote/archive rules) -> read `references/memory/memory-hygiene.md` 6. For the full canonical directory tree -> read `references/architecture/canonical-file-structure.md` 7. For the self-improving OS Triple-Loop and 3-file autoresearch