docs: restore DNA Memory core philosophy

README.md

@@ -19,6 +19,39 @@ DNA Memory 把一个 Markdown/Obsidian 目录作为长期记忆真源，把 SQLi
 它不会把“保存所有聊天记录”等同于“形成记忆”。完整会话仍由原客户端保管；
 DNA Memory 只保存有界来源指针、经过审查的提案，以及验证后的关键结论。
 
+## 核心理念：从对话数据到可复用认知
+
+> **DNA Memory 不把“保存对话”视为“形成记忆”。** 它受人类大脑记忆形成
+> 机制启发，把跨客户端的大量对话视为原始感知输入，通过信号提取、重要性与
+> 置信度加权、认知分型、验证结晶、召回反馈和新旧认知替代，持续把数据抽象为
+> 可复用认知。
+
+它不是另一个无限增长的聊天数据库。原始会话只是形成认知的材料；只有经过筛选、
+验证并值得在未来任务中复用的结论，才会进入长期记忆真源。
+
+```text
+大量原始对话
+  -> 有界采集与信号提取
+  -> 重要性、置信度与使用反馈加权
+  -> 认知分型
+  -> 验证后结晶为长期认知
+  -> 按当前任务召回
+  -> useful / misleading 反馈与 supersedes 演化
+```
+
+当前工程实现将长期认知分为四组、八种类型：
+
+| 认知模型 | 记忆类型 | 用途 |
+|---|---|---|
+| 个体倾向 | `preference` | 稳定偏好与协作习惯 |
+| 事实与洞察 | `fact`, `insight` | 已验证事实与从事实中抽象出的认识 |
+| 决策与状态 | `decision`, `project_state`, `open_loop` | 决策理由、当前状态和待闭环事项 |
+| 程序与经验 | `workflow`, `error_lesson` | 可复用流程和经过验证的失败经验 |
+
+这里的仿生学是设计启发，不是对人脑的完整复刻。DNA Memory 的价值不取决于
+保存了多少对话，而取决于形成了多少高质量认知，以及这些认知是否在后续决策中
+被准确召回和有效复用。
+
 ## 核心边界
 
 - Markdown 是长期真源，SQLite 是索引和遥测层。

README_EN.md

@@ -21,6 +21,44 @@ It does not equate storing every conversation with learning. Native transcripts
 stay in their original clients. DNA Memory stores bounded provenance pointers,
 reviewable proposals, and verified reusable conclusions.
 
+## Core philosophy: from conversation data to reusable cognition
+
+> **DNA Memory does not treat storing conversations as forming memories.** It is
+> a biomimetically inspired cognitive memory system. Cross-client conversations
+> are raw perceptual input; bounded collection, signal extraction, importance and
+> confidence weighting, cognitive classification, verification, recall feedback,
+> and explicit replacement progressively abstract that data into reusable
+> cognition.
+
+It is not another indefinitely growing chat database. Native conversations are
+source material; only conclusions that have been filtered, verified, and are
+worth reusing in future work become durable memory.
+
+```text
+raw conversations
+  -> bounded collection and signal extraction
+  -> importance, confidence, and usage-feedback weighting
+  -> cognitive classification
+  -> verified crystallization into durable cognition
+  -> task-focused recall
+  -> useful / misleading feedback and supersedes evolution
+```
+
+The current implementation organizes durable cognition into four groups and
+eight memory types:
+
+| Cognitive model | Memory types | Purpose |
+|---|---|---|
+| Individual tendencies | `preference` | Stable preferences and collaboration habits |
+| Facts and insights | `fact`, `insight` | Verified facts and knowledge abstracted from them |
+| Decisions and state | `decision`, `project_state`, `open_loop` | Rationale, current state, and unresolved work |
+| Procedures and experience | `workflow`, `error_lesson` | Reusable workflows and verified lessons from failure |
+
+Biomimetics is a design inspiration here, not a claim to replicate the human
+brain. DNA Memory's value is not how many conversations it stores, but how much
+high-quality cognition it forms and whether that cognition is accurately
+recalled and effectively reused in later decisions.
+
 ## Safety model
 
 - Markdown is durable truth; SQLite is search, queue, and telemetry state.

docs/superpowers/plans/2026-07-14-readme-core-philosophy.md

@@ -0,0 +1,200 @@
+# DNA Memory README Core Philosophy Implementation Plan
+
+> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
+
+**Goal:** Restore DNA Memory's biomimetically inspired cognitive-memory philosophy in the Chinese and English README files without overstating the implemented system.
+
+**Architecture:** Add a first-screen philosophy section to each README between the existing introduction and safety boundaries. The section maps the brand claim to current primitives: bounded collection, candidate extraction, importance/confidence/feedback ranking, eight durable memory types, recall, and explicit supersession.
+
+**Tech Stack:** Markdown, Python documentation safety checker, pytest, GitHub Actions
+
+---
+
+### Task 1: Add the Chinese core philosophy
+
+**Files:**
+- Modify: `README.md:19`
+
+- [ ] **Step 1: Insert the approved philosophy section**
+
+After the paragraph ending with “验证后的关键结论。”, insert:
+
+```markdown
+## 核心理念：从对话数据到可复用认知
+
+> **DNA Memory 不把“保存对话”视为“形成记忆”。** 它受人类大脑记忆形成
+> 机制启发，把跨客户端的大量对话视为原始感知输入，通过信号提取、重要性与
+> 置信度加权、认知分型、验证结晶、召回反馈和新旧认知替代，持续把数据抽象为
+> 可复用认知。
+
+它不是另一个无限增长的聊天数据库。原始会话只是形成认知的材料；只有经过筛选、
+验证并值得在未来任务中复用的结论，才会进入长期记忆真源。
+
+```text
+大量原始对话
+  -> 有界采集与信号提取
+  -> 重要性、置信度与使用反馈加权
+  -> 认知分型
+  -> 验证后结晶为长期认知
+  -> 按当前任务召回
+  -> useful / misleading 反馈与 supersedes 演化
+```
+
+当前工程实现将长期认知分为四组、八种类型：
+
+| 认知模型 | 记忆类型 | 用途 |
+|---|---|---|
+| 个体倾向 | `preference` | 稳定偏好与协作习惯 |
+| 事实与洞察 | `fact`, `insight` | 已验证事实与从事实中抽象出的认识 |
+| 决策与状态 | `decision`, `project_state`, `open_loop` | 决策理由、当前状态和待闭环事项 |
+| 程序与经验 | `workflow`, `error_lesson` | 可复用流程和经过验证的失败经验 |
+
+这里的仿生学是设计启发，不是对人脑的完整复刻。DNA Memory 的价值不取决于
+保存了多少对话，而取决于形成了多少高质量认知，以及这些认知是否在后续决策中
+被准确召回和有效复用。
+```
+
+- [ ] **Step 2: Verify the Chinese positioning is present**
+
+Run:
+
+```bash
+rg -n "## 核心理念：从对话数据到可复用认知|把数据抽象为|当前工程实现将长期认知分为四组、八种类型" README.md
+```
+
+Expected: three matches in the new first-screen section.
+
+### Task 2: Add the equivalent English philosophy
+
+**Files:**
+- Modify: `README_EN.md:20`
+
+- [ ] **Step 1: Insert the equivalent English section**
+
+After the paragraph ending with “verified reusable conclusions.”, insert:
+
+```markdown
+## Core philosophy: from conversation data to reusable cognition
+
+> **DNA Memory does not treat storing conversations as forming memories.** It is
+> a biomimetically inspired cognitive memory system. Cross-client conversations
+> are raw perceptual input; bounded collection, signal extraction, importance and
+> confidence weighting, cognitive classification, verification, recall feedback,
+> and explicit replacement progressively abstract that data into reusable
+> cognition.
+
+It is not another indefinitely growing chat database. Native conversations are
+source material; only conclusions that have been filtered, verified, and are
+worth reusing in future work become durable memory.
+
+```text
+raw conversations
+  -> bounded collection and signal extraction
+  -> importance, confidence, and usage-feedback weighting
+  -> cognitive classification
+  -> verified crystallization into durable cognition
+  -> task-focused recall
+  -> useful / misleading feedback and supersedes evolution
+```
+
+The current implementation organizes durable cognition into four groups and
+eight memory types:
+
+| Cognitive model | Memory types | Purpose |
+|---|---|---|
+| Individual tendencies | `preference` | Stable preferences and collaboration habits |
+| Facts and insights | `fact`, `insight` | Verified facts and knowledge abstracted from them |
+| Decisions and state | `decision`, `project_state`, `open_loop` | Rationale, current state, and unresolved work |
+| Procedures and experience | `workflow`, `error_lesson` | Reusable workflows and verified lessons from failure |
+
+Biomimetics is a design inspiration here, not a claim to replicate the human
+brain. DNA Memory's value is not how many conversations it stores, but how much
+high-quality cognition it forms and whether that cognition is accurately
+recalled and effectively reused in later decisions.
+```
+
+- [ ] **Step 2: Verify the English positioning is present**
+
+Run:
+
+```bash
+rg -n "## Core philosophy: from conversation data to reusable cognition|abstract that data into|four groups and" README_EN.md
+```
+
+Expected: three matches in the new first-screen section.
+
+### Task 3: Validate and publish the documentation change
+
+**Files:**
+- Verify: `README.md`
+- Verify: `README_EN.md`
+- Verify: `docs/superpowers/specs/2026-07-14-readme-core-philosophy-design.md`
+
+- [ ] **Step 1: Check terminology against implemented memory types**
+
+Run:
+
+```bash
+python3 - <<'PY'
+from scripts.markdown_memory import SUPPORTED_TYPES
+
+expected = {
+    "preference", "decision", "fact", "insight", "workflow",
+    "error_lesson", "project_state", "open_loop",
+}
+assert SUPPORTED_TYPES == expected
+print("README memory types match implementation")
+PY
+```
+
+Expected: `README memory types match implementation`.
+
+- [ ] **Step 2: Run the complete verification chain**
+
+Run:
+
+```bash
+python3 -m pytest -q
+python3 -m compileall -q dna.py dna scripts tests
+python3 scripts/check_public_safety.py
+git diff --check
+```
+
+Expected: 108 tests pass; compileall, public safety, and diff checks exit 0.
+
+- [ ] **Step 3: Commit the README implementation**
+
+```bash
+git add README.md README_EN.md
+GIT_COMMITTER_NAME="DNA Memory Contributors" \
+GIT_COMMITTER_EMAIL="noreply@users.noreply.github.com" \
+git commit --author="DNA Memory Contributors <noreply@users.noreply.github.com>" \
+  -m "docs: restore DNA Memory core philosophy"
+```
+
+Expected: one documentation commit with sanitized author and committer metadata.
+
+- [ ] **Step 4: Push, open a pull request, and verify GitHub Actions**
+
+```bash
+git push -u origin codex/readme-core-philosophy
+gh pr create --repo AIPMAndy/dna-memory \
+  --base main \
+  --head codex/readme-core-philosophy \
+  --title "docs: restore DNA Memory core philosophy" \
+  --body "Restore the biomimetically inspired cognitive-memory philosophy in the Chinese and English README files, mapped to the current implementation and safety boundaries."
+PR_NUMBER=$(gh pr view --repo AIPMAndy/dna-memory --json number --jq .number)
+gh pr checks --watch --fail-fast "$PR_NUMBER" --repo AIPMAndy/dna-memory
+```
+
+Expected: the PR is mergeable and all required checks pass.
+
+- [ ] **Step 5: Merge and verify the public main branch**
+
+```bash
+PR_NUMBER=$(gh pr view --repo AIPMAndy/dna-memory --json number --jq .number)
+gh pr merge "$PR_NUMBER" --repo AIPMAndy/dna-memory --squash --delete-branch
+gh pr view "$PR_NUMBER" --repo AIPMAndy/dna-memory --json state,mergedAt,mergeCommit
+```
+
+Expected: state is `MERGED`; the public `main` README files contain the new philosophy sections.

docs/superpowers/specs/2026-07-14-readme-core-philosophy-design.md

@@ -0,0 +1,80 @@
+# DNA Memory README 核心理念设计
+
+## 目标
+
+在中英文 README 的首屏介绍中恢复 DNA Memory 的核心定位：它不是一个单纯保存
+对话的数据仓库，而是一个受人类大脑记忆形成机制启发的认知提炼系统。README 需要
+同时具备鲜明的品牌表达和可核验的工程依据。
+
+## 核心定位
+
+中文主张：
+
+> DNA Memory 不把“保存对话”视为“形成记忆”。它受人类大脑记忆机制启发，
+> 把跨客户端的大量对话视为原始感知输入，通过信号提取、重要性与置信度加权、
+> 认知分型、验证结晶、召回反馈和新旧认知替代，持续把数据抽象为可复用认知。
+
+英文表达保持同一含义，使用 `biomimetically inspired cognitive memory system`，避免使用
+`brain replica` 等无法由当前实现支持的表述。
+
+## README 结构
+
+在现有项目简介之后、核心边界或安全模型之前，新增“核心理念 / Core philosophy”章节。
+章节由四部分组成：
+
+1. 一段品牌宣言，说明原始对话、长期记忆和可复用认知的区别。
+2. 一条从对话数据到长期认知的处理链路。
+3. 一组与当前八种记忆类型一致的认知分型。
+4. 三条工程边界，防止将仿生学启发误读为完整复刻人脑。
+
+## 认知处理链路
+
+README 使用以下流程表达系统价值：
+
+```text
+大量原始对话
+  -> 有界采集与信号提取
+  -> 重要性、置信度与使用反馈加权
+  -> 认知分型
+  -> 验证后结晶为长期认知
+  -> 按当前任务召回
+  -> useful / misleading 反馈与 supersedes 演化
+```
+
+该流程映射当前实现中的原生历史导入、候选提案、`importance`、`confidence`、
+`memory_feedback`、Markdown 长期真源和 `supersedes`，不引入新的运行时行为。
+
+## 认知分型
+
+README 将现有八种类型组织为四组认知模型，底层类型名保持不变：
+
+| 认知模型 | 现有类型 | 含义 |
+|---|---|---|
+| 个体倾向 | `preference` | 稳定偏好与协作习惯 |
+| 事实与洞察 | `fact`, `insight` | 已验证事实与从事实中抽象出的认识 |
+| 决策与状态 | `decision`, `project_state`, `open_loop` | 决策理由、当前状态和待闭环事项 |
+| 程序与经验 | `workflow`, `error_lesson` | 可复用流程和经过验证的失败经验 |
+
+这里的“模型”是工程上的认知分型，不宣称与神经科学中的工作记忆、情景记忆、
+语义记忆或程序记忆一一对应。
+
+## 工程边界
+
+- 仿生学是设计启发，不是对人脑的完整模拟。
+- 原始会话继续留在客户端；DNA Memory 保存来源指针、候选和验证后的认知结论。
+- 系统价值不以保存量衡量，而以高质量认知的形成、召回和后续有效复用衡量。
+
+## 修改范围
+
+- 修改 `README.md`，增加中文核心理念、链路和认知分型。
+- 修改 `README_EN.md`，同步相同的英文定位。
+- 保留现有安装、MCP、运维、安全和公开发布说明。
+- 不修改 Python 代码、数据结构、MCP 接口或用户运行数据。
+
+## 验收标准
+
+1. 中英文 README 首屏都明确出现仿生学启发和“从数据抽象出认知”的定位。
+2. 文案明确区分原始对话、候选记忆和验证后的长期认知。
+3. 所有工程能力均可映射到当前代码和公开接口。
+4. 中英文叙事含义一致，不包含个人配置或私人数据。
+5. `python3 scripts/check_public_safety.py` 与 `git diff --check` 通过。