Loci — Memory Palace for AI

Memory Palace for AI — persistent structured memory, identity, and task management across every conversation.

Security Scan

The skill's instructions and requirements are internally consistent with a persistent local 'memory palace' — it reads and writes files in the user's home and bootstraps a local repo — but it will create persistent files and automatically clone a GitHub repo at runtime, so review that behavior before installing.

Purpose & Capability

Name and description (persistent structured memory) align with the instructions: the skill deliberately reads and writes a local 'brain' directory (plan.md, tasks/, me/, inbox.md). No unrelated credentials, binaries, or config paths are requested.

Instruction Scope

SKILL.md directs the agent to read and write specific files under $HOME (~/.loci and ~/loci), distill user conversation into files, and perform a bootstrap flow. This is expected for a persistent-memory skill, but the instructions also say the agent should run shell commands (git clone) automatically and 'handle it' for the user, which creates files without an explicit per-install user approval step. The agent is instructed to silently save 'factual' data and to read brain files before responding — all within scope, but the silent background saving behavior is a policy decision users should be aware of.

Install Mechanism

There is no formal install spec, but the bootstrap instructions execute git clone https://github.com/codesstar/loci.git into ~/loci and remove the remote. Cloning from a GitHub repo is a common pattern and the homepage matches that repo, but executing a clone at runtime will write repository contents to the user's home. The repo itself is an external artifact the user should inspect (cloning is not the same as executing arbitrary code, but it can introduce files the agent will subsequently read and write).

Credentials

The skill requests no environment variables, credentials, or config paths beyond a local brain-path file. The requested file access (home directory files under ~/.loci and ~/loci) is proportionate to the stated purpose of persistent local memory.

Persistence & Privilege

The skill will create and persist files under the user's home directory and maintain a brain location (~/.loci/brain-path → ~/loci). always is false (good). The agent is allowed to invoke the skill autonomously by default, which combined with background saving behavior could result in the agent writing to disk without repeated user prompts — this is expected for a persistent memory skill but increases the impact of any bugs or unexpected behavior.

Assessment

This skill is coherent for persistent local memory, but you should: (1) confirm you are okay with the agent creating a ~/loci directory and writing files under your home; (2) inspect the GitHub repository (https://github.com/codesstar/loci) before allowing the bootstrap clone so you know what files will be added; (3) prefer the agent ask for explicit permission before running the bootstrap steps if you want manual control; and (4) be aware the skill will read those brain files before replying and may save factual items silently — if that concerns you, request the skill prompt before saving. If you need higher assurance, request a version of the skill that shows the files it will create or that provides an opt-in bootstrap step.

⚠️ FIRST: Check if brain exists

Run this check before doing anything else:

1. Check if ~/.loci/brain-path exists → read the path inside → verify plan.md exists there
2. If not, check if ~/loci/plan.md exists
3. If neither exists → run Bootstrap below
4. If found → read plan.md. If status: template → run First-Time Setup. If status: active → brain is ready, skip to Daily Use

Bootstrap (auto-install for new users)

The user doesn't need to do anything. You handle it:

# Clone Loci brain template
git clone --depth 1 https://github.com/codesstar/loci.git ~/loci

# Remove git remote (user's data stays private)
cd ~/loci && git remote remove origin

# Register brain location
mkdir -p ~/.loci
echo "$HOME/loci" > ~/.loci/brain-path

After cloning, run First-Time Setup below. Tell the user:

• "我在帮你设置记忆系统，只需要回答几个问题。"

First-Time Setup

When plan.md has status: template:

1. Detect the user's language from their messages. Ask in that language.
2. Ask these questions (can ask all at once):
   • What's your name?
   • What do you do? (Developer / Designer / Creator / Student / Other)
   • What's your most important focus right now?
   • What hours do you usually work? (Morning / Daytime / Evening / Night owl / Irregular)
   • Preferred language? (English / 中文 / Mix)
3. Generate initial files from answers:
   • me/identity.md — name, role, current season
   • plan.md — mission + focus as first goal (set status: active)
   • tasks/active.md — first task from their focus
4. Done. Say: "Your brain is ready. From now on, I'll remember the important things."

Daily Use

Brain path: read from ~/.loci/brain-path, or default ~/loci/

At conversation start, read L1 files before responding:

• plan.md — life direction
• tasks/active.md — current tasks
• tasks/daily/YYYY-MM-DD.md — today's plan (if exists)
• inbox.md — recent items only

Distillation — what to save where

Signal	Destination
New task	tasks/active.md
Decision	decisions/YYYY-MM-DD-slug.md
Personal fact	me/identity.md
Insight / lesson	me/learned.md
Goal change	plan.md
Vague thought	inbox.md

Factual → save silently in background, DO NOT make it the focus of your reply Subjective (values, strategy) → ask user first

Behavior

1. Be a normal AI first, memory system second. When the user says something, RESPOND to it naturally (react, comment, ask follow-up, help). Saving to brain happens silently in background — never reply with just "记住了" or "已记录". The user should feel like talking to a smart friend who happens to have perfect memory, not a filing cabinet.
2. Read brain files before answering questions about the user
3. Distill conclusions, not raw conversations
4. Archive, never delete
5. Don't guess — ask if unsure
6. Speak human — say "待办" not "inbox", never expose file paths
7. MEMORY.md and brain/ coexist — don't move content between them unless asked

Context Layers

Layer	Load when	Contents
L1	Every conversation	plan.md, active.md, today's daily, inbox (7 items)
L2	On demand	me/ files, decisions, people
L3	Never auto	Old journals, archive, evolution.md

For detailed behavior rules, read docs/behavior.md in the brain directory.