memory-bootstraplisted

# Memory Bootstrap **This is the single skill every other skill in this system depends on.** If a session doesn't start by reading project memory, nothing downstream (recall, checkpointing, handoff) has anything to build on — this is the entry point of the whole architecture described in [ARCHITECTURE.md](../../ARCHITECTURE.md) of the `agent-memory-protocol` repo. ## What to read, in order, at session start 1. **`.memory/BOOTSTRAP.md`** — the compact (<150 line) orientation: what this project is, current architecture summary, active conventions, non-negotiables. This is the equivalent of a coding agent's `AGENTS.md`/`CLAUDE.md`, scoped specifically to memory-system orientation. 2. **`.memory/HANDOFF.md`** — the most recent explicit handoff note, if one exists (written by `agent-handoff` when a prior session deliberately switched tools/agents). 3. **The most recent file in `.memory/checkpoints/`** — sorted by timestamp, this is the actual last-known state, which matters most specifically because a session might have ended *without* a clean handoff (credit exhaustion, crash) — see `session-checkpoint`'s rationale for why this file, not `HANDOFF.md`, is the authoritative "where things actually stood." 4. **Any `.memory/tasks/*.md` files marked active** — current in-flight work, so the new session doesn't have to be told what's being worked on. ## What this produces A working understanding of: what the project is, what was being worked on, what decisions have already been ma