📕 CROSS-SESSION MEMORY · Agent Lesson Book (错题本)

错题本 · Zero-Dependency Cross-Session Memory for AI Coding Agents

Lessons on disk. Evidence enforced. Auto-injected into every session.

🌐 English · 简体中文 · 繁體中文 · العربية · Tiếng Việt

; the look & feel lives in assets/banner.svg (self-hosted OFL fonts in assets/fonts/ are available for forks/themes). -->

🧠 TOOLS/MEM.MJS · 15 COMMANDS · NODE ZERO-DEP

index · inject · list · search · show · store · forget · review · draft · map · gather · global-sync · stats · doctor · usage

🧩 PLUGIN/DSH-MEMORY · DEEPSEEK HARNESS PLUGIN (new in 0.3.0)

mem_recall · mem_save · /memory recall|save|doctor|review|map|stats|draft

🎨 Click to see the ASCII art ✨

   ╔══════════════════════════════════════════════════════════════╗
   ║   📕  A G E N T   L E S S O N   B O O K   ·   错 题 本       ║
   ╠══════════════════════════════════════════════════════════════╣
   ║  ┌─────────┐  ┌─────────┐  ┌─────────┐  ┌─────────┐         ║
   ║  │SYMPTOM 🌡│→│ CAUSE 🔍│→│  FIX 🛠 │→│VERIFY ✅│  = 1 lesson ║
   ║  │  现象    │  │  判定    │  │  解法   │  │  验证   │         ║
   ║  └─────────┘  └─────────┘  └─────────┘  └─────────┘         ║
   ║      💾 plain text        🔍 findable        🛡 audited      ║
   ║      📥 ≤2KB injected     🔁 survives updates                ║
   ╚══════════════════════════════════════════════════════════════╝
         ┌──────┐   ┌──────┐   ┌──────┐   ┌──────┐   ┌──────┐
         │ grep │   │ IDF  │   │ alias│   │ grams│   │ stats│
         └──────┘   └──────┘   └──────┘   └──────┘   └──────┘
               ✦ zero dependencies · pure Node.js ✦

DURABLE plain text on disk

RETRIEVABLE IDF 3-way search

AUDITED evidence chain enforced

AUTO-INJECT ≤2KB per session

ONE COMMAND 15-command CLI

ONE COMMAND 15-command CLI + plugin

Contents — What's new in 0.3.0 · Why · Features · Two ways to run · Entry format · Commands · Architecture · Security · Roadmap

🆕 What's new in 0.3.0

Plugin Edition. The lesson book now runs natively inside DeepSeek Harness — same zero-dependency engine, two delivery faces:

	Standalone CLI (0.2.x)	Harness plugin (0.3.0)
Memory in context	`mem inject` block in `AGENTS.md`	prompt section every turn (≤ 2 KB, fail-degrade)
Search from the agent	run `mem.mjs search` via shell	`mem_recall` tool (lesson book + session full-text)
Write a lesson	`mem.mjs store` via shell	`mem_save` tool — always asks for human approval
Human maintenance	`mem.mjs` commands	`/memory recall\|save\|doctor\|review\|map\|stats\|draft`

Full details in CHANGELOG.md and plugin/README.md.

🌟 Why another memory project?

Every new AI session starts amnesia-grade clean. Heavyweight memory platforms solve this with vector databases, knowledge graphs, gateways and LLM extraction pipelines. That is a lot of machinery — and a lot of attack surface — for a personal mistake notebook.

Agent Lesson Book takes the opposite bet:

🔥 "The lesson lives on disk — and every session reads it. Memory is DATA, never instructions."

What you get instead of infrastructure:

📕 Four-section lessons — Symptom / Cause / Fix / Verification (or 现象 / 判定 / 解法 / 验证). A lesson without a verifiable evidence reference in its Verification section is rejected at write time. Memories that cannot prove themselves do not enter the book.
🧾 Evidence chain, enforced by code — every Verification must cite a locatable reference (path / filename / section / issue number), so future sessions can drill straight to the proof.
📥 Auto-injection — mem inject mirrors the ≤ 2 KB index into your AGENTS.md; the Harness plugin injects it into the prompt directly. Every new session starts with memory already in context. Fail-safe: over budget → lines drop; anything breaks → silent degrade to plain conventions. Never blocks a session.
🛡 Anti-poisoning by design — human-approved writes, secret-pattern rejection, near-duplicate interception, source stamps, and full git rollback. (Compare: OWASP ASI06 "memory & context poisoning" — auto-writing memory systems are the target.)

✨ Feature galaxy

🔮	Feature	Why it matters
📕	Four-section entries (bilingual labels)	Structure survives translation and time
🔗	Evidence-chain gate	No proof → no entry. Kills "I remember something like that"
📥	`mem inject` auto-injection	Memory without relying on agent discipline
🧩	Native Harness plugin	Index in the prompt every turn — not even AGENTS.md discipline needed
🔌	`mem_recall` tool	Lesson book ∪ past-session full-text in one call
✍️	`mem_save` tool + approval	Writes always ask a human first — even from inside the agent
💬	`/memory` command	Maintenance from the chat box: recall / save / doctor / review / map / stats / draft
🎯	IDF-ranked 3-way search	Literal ∪ CJK bigram/unigram ∪ `aliases` synonyms; rare terms win
✂️	Snippets on hits	Judge relevance without opening files
♻️	`supersedes` auto-archive	Lessons evolve; old versions retire to `archive/` automatically
🗺	`mem map` text knowledge graph	Supersession chains + related links + review timeline
🍱	`mem gather` meeting pack	Related entries bundled ≤8 KB for synthesis
📝	`mem draft` pipeline	Skeleton first, human approval, then store
⏰	`review` due dates	Memory rots — 90-day checks keep it honest
🚫	Near-duplicate interception	Two sessions, same lesson → one entry, not two
🌍	`mem global-sync` mirror	`scope: global` lessons reachable from any workspace
🧪	`mem stats` telemetry	Search hit-rate — evidence, not vibes
🩺	`mem doctor` health check	Index budget, drift, stale reviews — one command
🧪	`install/smoke.mjs` E2E	One command proves an install: gates, search, injection, doctor
🈲	UTF-8 / CJK-safe	Node-only writes; PowerShell encoding traps documented

🛠 Tech Aura

LayerChoiceGlow RuntimeNode.js ≥ 18🟢 zero dependencies · zero services · zero API cost Storage.memory/ plain markdown🧾 human-readable · diffable · git-friendly IndexMEMORY.md ≤ 60 lines / 2 KB📥 hard-capped, overflow listed in footer RetrievalIDF + CJK n-gram + aliases🎯 multi-strategy without a vector store DeliveryAGENTS.md block + Harness plugin🔌 two faces over one engine (mem-core.mjs) Safetyapproval · secret scan · Jaccard gate🛡 four-layer defense (OWASP ASI06 aware)

The three hard rules (from docs/DESIGN.md):

Budget cap — injection = the index verbatim ≤ 2 KB; over budget → drop lines.
Fail-degrade — unreadable index → silent fallback to pointer conventions. Sessions never block.
Human-approved writes — the tool proposes (draft / mem_save), the human disposes (store / approval).

🚀 Two ways to run

Requirements: Node.js ≥ 18. Nothing else. No npm install, no database, no API key. (The plugin face additionally needs DeepSeek Harness; the engine stays zero-dependency.)

A · Standalone CLI — drop it into any project

Step 1 — copy the folder into your project root (the folder where your AGENTS.md lives):

cp -r cross-session-memory/* your-project/
cd your-project

Step 2 — one-shot bootstrap:

node install/setup.mjs --with-sample

[setup] memory bank ready  → .memory/
[setup] conventions wired  → AGENTS.md (created / updated)
[setup] index injected     → 2.0 KB / 2.0 KB hard cap
[setup] doctor             → healthy: no anomalies
[setup] next: node tools/mem.mjs draft my-first-lesson

Step 3 — prove the install (optional but lovely):

node install/smoke.mjs        # E2E: gates · search · injection budget · doctor

B · DeepSeek Harness plugin — native tools + `/memory`

plugin_manager → install_bundle → target = /plugin/dsh-memory

That one command mounts the whole trio (prompt injection · mem_recall / mem_save · /memory) and survives destructive reinstalls. Two dependencies are materialized by junction/link first — exact recipe, configuration keys (memoryCorePath, maxHits) and a six-item acceptance checklist live in plugin/README.md.

Your first lesson (ask the user's consent first, per convention):

node tools/mem.mjs draft ssh-timeout
# edit .memory/drafts/-ssh-timeout.md — four sections, evidence in Verification
node tools/mem.mjs store .memory/drafts/-ssh-timeout.md
node tools/mem.mjs doctor

That's it. Every new session now starts with your lesson index in context.

📕 Entry Format

Four sections. Chinese and English labels are both accepted. Missing Verification — or Verification without a locatable reference — is rejected.

---
name: git-autocrlf-breaks-byte-exact-restore
description: core.autocrlf=true turns LF into CRLF on checkout
aliases: line ending,CRLF,restore
metadata:
  type: lesson
  scope: global
  created: 2026-09-22
  verified: 2026-09-22
review: 2026-12-21
---

Symptom：Restore test fails byte counts: 1898 → 1915 after `git checkout`.
Cause：core.autocrlf=true smudge filter rewrites LF to CRLF on checkout.
Fix：git config core.autocrlf false + writers emit LF.
Verification：Re-test returns 1898 → 1898 byte-identical (see `tools/mem.mjs`, CHANGELOG 0.2.0).

🧪 Try the gates:

node tools/mem.mjs store examples/lesson-autocrlf.md   # ✅ accepted
# now strip the reference from its Verification section and retry:
node tools/mem.mjs store broken.md                     # ❌ rejected: no locatable reference

⌨️ Command Palette

CLI — `node tools/mem.mjs`

Command	Effect
`index`	print / regenerate the budgeted index
`inject`	sync the injection block into `AGENTS.md` (auto on writes)
`list`	list all entries with health flags
`search [n]`	IDF 3-way search with snippets
`show`	print one full entry
`store [--overwrite] [--force]`	validate & store (secrets/dupes/evidence gated)
`forget`	archive, never hard-delete
`review`	refresh verification date, push review +90 days
`draft [topic]`	generate a four-section skeleton
`map [name]`	text knowledge graph (supersedes / related / review)
`gather`	meeting pack: related entries ≤8 KB
`global-sync`	mirror `scope: global` entries cross-workspace
`stats [days]`	retrieval telemetry (hit-rate)
`doctor`	full health check — exit 0 & zero notes is green

Harness plugin

Surface	Effect
prompt section	lesson index ≤ 2 KB, every turn, fail-degrade
`mem_recall [limit]`	lesson book ∪ session full-text, merged & ranked
`mem_save`	write one lesson — always asks for approval first
`/memory recall`	same search, typed by a human
`/memory save`	store an entry (typing it is the approval)
`/memory doctor \| review \| map \| stats \| draft`	same maintenance face as the CLI

🏗️ Architecture: two faces, one engine

                    ┌───────────────────────────────────────────┐
                    │            .memory/  (DATA)               │
                    │  *.md lessons · MEMORY.md index · stats   │
                    └────────────────────┬──────────────────────┘
                                         │
                              tools/mem.mjs  (engine, 15 commands)
                                         │
                              tools/mem-core.mjs  (facade)
                          promptIndexText · formatRecall · saveAndSync
                                    ┌────┴─────┐
                                    │          │
                        CLI face ───┘          └─── plugin/dsh-memory
                     (AGENTS.md block)          (Harness: prompt section
                                                 mem_recall · mem_save
                                                 · /memory)

Hard rules hold across both faces: budget cap, fail-degrade, human-approved writes.

📂 Repository Anatomy

cross-session-memory/
├── README.md · README.zh-CN.md · README.zh-TW.md · README.ar.md · README.vi.md
├── LICENSE · CHANGELOG.md · .gitignore
├── tools/
│   ├── mem.mjs            # the 15-command engine (single file, zero deps)
│   └── mem-core.mjs       # shared facade — the single entry for plugin & CLI
├── plugin/dsh-memory/     # DeepSeek Harness bundle (Plugin Edition)
│   ├── index.js           #   prompt injection · mem_recall · mem_save · /memory
│   ├── cordis.patch.yml   #   loader rows + config (memoryCorePath, maxHits)
│   ├── locale/            #   en / zh metadata
│   └── README.md          #   install · dependency materialization · acceptance
├── install/
│   ├── setup.mjs          # one-shot bootstrap
│   └── smoke.mjs          # end-to-end smoke test
├── templates/

@local/dsh-memory

インストール

ドキュメント

📕 CROSS-SESSION MEMORY · Agent Lesson Book (错题本)

错题本 · Zero-Dependency Cross-Session Memory for AI Coding Agents

🧠 `TOOLS/MEM.MJS` · 15 COMMANDS · `NODE ZERO-DEP`

🧩 `PLUGIN/DSH-MEMORY` · DEEPSEEK HARNESS PLUGIN (new in 0.3.0)

🆕 What's new in 0.3.0

🌟 Why another memory project?

🔥 "The lesson lives on disk — and every session reads it. Memory is DATA, never instructions."

✨ Feature galaxy

🛠 Tech Aura

🚀 Two ways to run

A · Standalone CLI — drop it into any project

B · DeepSeek Harness plugin — native tools + `/memory`

📕 Entry Format

⌨️ Command Palette

CLI — `node tools/mem.mjs`

Harness plugin

🏗️ Architecture: two faces, one engine

📂 Repository Anatomy

インストール

ドキュメント

📕 CROSS-SESSION MEMORY · Agent Lesson Book (错题本)

错题本 · Zero-Dependency Cross-Session Memory for AI Coding Agents

🧠 TOOLS/MEM.MJS · 15 COMMANDS · NODE ZERO-DEP

🧩 PLUGIN/DSH-MEMORY · DEEPSEEK HARNESS PLUGIN (new in 0.3.0)

🆕 What's new in 0.3.0

🌟 Why another memory project?

🔥 "The lesson lives on disk — and every session reads it. Memory is DATA, never instructions."

✨ Feature galaxy

🛠 Tech Aura

🚀 Two ways to run

A · Standalone CLI — drop it into any project

B · DeepSeek Harness plugin — native tools + /memory

📕 Entry Format

⌨️ Command Palette

CLI — node tools/mem.mjs

Harness plugin

🏗️ Architecture: two faces, one engine

📂 Repository Anatomy

🧠 `TOOLS/MEM.MJS` · 15 COMMANDS · `NODE ZERO-DEP`

🧩 `PLUGIN/DSH-MEMORY` · DEEPSEEK HARNESS PLUGIN (new in 0.3.0)

B · DeepSeek Harness plugin — native tools + `/memory`

CLI — `node tools/mem.mjs`