memorytrace 0.1.0__tar.gz → 0.3.0__tar.gz

memorytrace-0.3.0/.claude/workspace/evaluation.md

@@ -0,0 +1,72 @@
+# Evaluation: Engram 3-Session Vibe Coding E2E Simulation
+
+## Verdict: PASS
+
+## Test Results Summary
+
+| Step | Test | Status | Details |
+|------|------|--------|---------|
+| Session 1 - Save #1 | Tech stack save | PASS | Saved successfully, no entities auto-detected |
+| Session 1 - Save #2 | Church model save | PASS | Saved successfully |
+| Session 1 - Save #3 | JWT auth save | PASS | Saved, entities detected: Access Token, Refresh Token |
+| Session 1 - Save #4 | Kakao map save | PASS | Saved successfully |
+| Session 1 - Save #5 | Session summary save | PASS | Saved successfully |
+| Session 2 - Save #1 | Review model save | PASS | Saved successfully |
+| Session 2 - Save #2 | CORS fix save | PASS | Saved successfully |
+| Session 2 - Save #3 | PostGIS search save | PASS | Saved successfully |
+| Session 2 - Save #4 | Session summary save | PASS | Saved successfully |
+| Session 3 - recent 10 | All 9 notes returned | PASS | All 9 notes listed in reverse chronological order |
+| Session 3 - find JWT | JWT note found | PASS | Returned exact match |
+| Session 3 - find CORS | CORS note found | PASS | Returned exact match |
+| Session 3 - find Review | Review model found | PASS | Returned exact match |
+| Session 3 - find PostGIS | PostGIS notes found | PASS | Returned 3 matching notes (tech stack, PostGIS specific, Church model) |
+| Session 3 - status | Correct counts | PASS | 2 entities, 0 facts, 9 notes, 0 conflicts, healthy |
+| Dedup - duplicate save | No new note created | PASS | Output said "Updated" (not "Saved"), notes stayed at 9 |
+| Delete - preview | Shows matching note | PASS | Found 1 note, displayed preview, showed confirm command |
+| Delete - confirm | Note deleted | PASS | "Deleted 1 note(s)" |
+| Delete - verify gone | Search returns nothing | PASS | "Nothing found for CORS" |
+| Delete - count check | Notes dropped to 8 | PASS | Status shows 8 notes |
+
+## Criteria Results
+
+| Criterion | Status | Notes |
+|-----------|--------|-------|
+| Multi-session persistence | PASS | Notes saved in session 1 and 2 fully retrievable in session 3 |
+| Korean text handling | PASS | All Korean content stored and retrieved without corruption |
+| Search accuracy | PASS | All 4 search queries returned correct results |
+| Deduplication | PASS | Exact duplicate did not increment notes count (stayed at 9) |
+| Delete preview (safe mode) | PASS | Without --confirm, shows preview only and prompts for confirmation |
+| Delete execution | PASS | With --confirm, deletes the note and confirms count |
+| Delete verification | PASS | Deleted note no longer appears in search |
+| Entity auto-detection | PASS | Detected "Access Token" and "Refresh Token" as entities from JWT note |
+| Status reporting | PASS | Accurate counts at every checkpoint |
+
+## Issues Found
+
+### [Issue 1 - Severity: Low]
+- **What**: Entity auto-detection is sparse -- only 2 entities detected across 9 notes containing terms like "NestJS", "Prisma", "PostgreSQL", "PostGIS", "Church", "Review", "JWT", "CORS"
+- **Where**: Entity extraction logic (internal to engram)
+- **Expected**: More domain terms would be recognized as entities
+- **Actual**: Only "Access Token" and "Refresh Token" were detected (2 entities total)
+- **Fix suggestion**: This is a minor observation. The entity detection appears to focus on capitalized multi-word terms. Not a functional problem since `find` search works correctly regardless.
+
+### [Issue 2 - Severity: Low]
+- **What**: `engram recent 10` requested 10 notes but only 9 existed -- no error or warning, just returned all 9
+- **Where**: `recent` subcommand
+- **Expected**: This is actually correct behavior (return up to N)
+- **Actual**: Correct -- returned all 9 without error
+- **Fix suggestion**: No fix needed, this is the expected behavior.
+
+## Regressions
+
+None detected. All core operations (save, find, recent, status, delete-note, dedup) function correctly.
+
+## Summary
+
+All 20 test steps passed. The engram tool correctly handles:
+- Persistent storage across simulated sessions
+- Korean (UTF-8) content without corruption
+- Full-text search returning accurate results
+- Deduplication of exact-match notes (count stays stable)
+- Two-phase delete (preview then confirm) with verification
+- Accurate status reporting at every checkpoint

{memorytrace-0.1.0 → memorytrace-0.3.0}/.gitignore

@@ -1,20 +1,20 @@
-__pycache__/
-*.py[cod]
-*.egg-info/
-dist/
-build/
-.eggs/
-*.egg
-.pytest_cache/
-.mypy_cache/
-.ruff_cache/
-*.db
-*.db-wal
-*.db-shm
-.engram/
-/tmp/
-*.tmp
-.DS_Store
-.env
-.venv/
-venv/
+__pycache__/
+*.py[cod]
+*.egg-info/
+dist/
+build/
+.eggs/
+*.egg
+.pytest_cache/
+.mypy_cache/
+.ruff_cache/
+*.db
+*.db-wal
+*.db-shm
+.engram/
+/tmp/
+*.tmp
+.DS_Store
+.env
+.venv/
+venv/

{memorytrace-0.1.0 → memorytrace-0.3.0}/CLAUDE.md

@@ -90,6 +90,19 @@ engine.close()
 - 코어 엔진에 `print()` 금지 — CLI 계층에서만 출력
 - 모든 메서드는 구조화된 객체를 반환 (stdout 오염 없음)
 - 타입 힌트 필수, `from __future__ import annotations` 사용
+## Dependencies
+
+- **jq** (필수): 모든 Claude Code hooks가 JSON 파싱에 사용. 미설치 시 hooks가 조용히 스킵됨
+  - Windows: `winget install jqlang.jq`
+  - macOS: `brew install jq`
+  - Linux: `apt install jq` / `yum install jq`
+- **engram** (선택): AI 대화 간 지식 유지 메모리. `pip install memorytrace`
+
+## Windows 호환성
+
+- Windows에서 `engram` CLI 사용 시 `PYTHONIOENCODING=utf-8` 환경변수 필요
+- `cli_entry()`에서 stdout/stderr를 UTF-8로 reconfigure하여 cp949 인코딩 에러 방지
+- 상세: `docs/05-windows-compatibility/windows-encoding.md` 참고
 
 ## Design Principles
 1. **에이전트 퍼스트**: MCP/SDK가 1급 인터페이스, CLI는 래퍼

{memorytrace-0.1.0 → memorytrace-0.3.0}/LICENSE

@@ -1,21 +1,21 @@
-MIT License
-
-Copyright (c) 2026 Engram Contributors
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
+MIT License
+
+Copyright (c) 2026 Engram Contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

{memorytrace-0.1.0 → memorytrace-0.3.0}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.4
 Name: memorytrace
-Version: 0.1.0
+Version: 0.3.0
 Summary: AI agent memory system with SQLite+FTS5, MCP integration, and quality gates
-Project-URL: Homepage, https://github.com/engram-memory/engram
-Project-URL: Issues, https://github.com/engram-memory/engram/issues
+Project-URL: Homepage, https://github.com/aop60003/default
+Project-URL: Issues, https://github.com/aop60003/default/issues
 License-Expression: MIT
 License-File: LICENSE
 Keywords: agent,ai,fts5,llm,mcp,memory,sqlite
@@ -43,7 +43,7 @@ AI agent memory that learns across conversations.
 ## Install
 
 ```bash
-pip install git+https://github.com/aop60003/default.git
+pip install memorytrace
 ```
 
 ## Use

{memorytrace-0.1.0 → memorytrace-0.3.0}/README.md

@@ -1,100 +1,100 @@
-# Engram
-
-AI agent memory that learns across conversations.
-
-## Install
-
-```bash
-pip install git+https://github.com/aop60003/default.git
-```
-
-## Use
-
-```bash
-engram save "Minseong Jeong is the Space King of Galaxy Corp"
-engram find "Space King"
-engram who "Minseong Jeong"
-```
-
-That's it. 3 commands. No config, no API keys, no database setup.
-
-## All Commands
-
-```
-engram save "text"                   Save information (auto-extracts entities & facts)
-engram find "query"                  Search memory
-engram who "name"                    Look up a person/org
-engram remember "name" "fact"        Manually remember something
-engram all                           List everything
-engram status                        Health check
-engram forget "name"                 Delete an entity
-engram export                        Export to Markdown files
-```
-
-## Examples
-
-```bash
-# Store meeting notes
-engram save "Met Alice Johnson, VP of Engineering at Google. Discussed cloud AI partnership."
-
-# Store Korean
-engram remember "정민성" "AI와 우주 탐사에 관심이 많다"
-
-# Search later
-engram find "Google partnership"
-engram who "Alice Johnson"
-
-# Check status
-engram status
-```
-
-## What It Does Automatically
-
-- **Extracts** people & organizations from text (English, Korean, Chinese, Japanese)
-- **Masks PII** — phone numbers, credit cards, emails become `[REDACTED]`
-- **Deduplicates** — same fact won't be stored twice
-- **Detects conflicts** — "CEO" then "CTO" for same person gets flagged
-- **Generates summaries** — auto-summary when you don't provide one
-- **Learns across sessions** — every conversation builds on previous ones
-
-## For Claude Code Users
-
-Add slash commands to use memory inside Claude Code:
-
-```bash
-# Copy command files
-mkdir -p ~/.claude/commands
-# See docs/04-usage-guide/02-claude-code-setup.md for details
-```
-
-Then use: `/memory-save`, `/memory-find`, `/memory-who`, `/memory-remember`, `/memory-status`
-
-## For Developers (Python SDK)
-
-```python
-from engram.integrations.sdk import EngramSDK
-
-with EngramSDK() as sdk:
-    sdk.store("Minseong Jeong is the Space King of Galaxy Corp")
-    result = sdk.search("Space King")
-    print(result.to_agent_context())
-```
-
-## How It Works
-
-```
-Text → NER Extraction → Quality Gate → SQLite+FTS5 → BM25 Search
-                              ↓
-                     Confidence Scoring
-                     PII Masking
-                     Duplicate Check
-                     Conflict Detection
-```
-
-- **Storage**: SQLite + FTS5 (zero dependencies, genuine BM25)
-- **Export**: Markdown files for human reading
-- **Session**: Context carries over between conversations
-
-## License
-
-MIT
+# Engram
+
+AI agent memory that learns across conversations.
+
+## Install
+
+```bash
+pip install memorytrace
+```
+
+## Use
+
+```bash
+engram save "Minseong Jeong is the Space King of Galaxy Corp"
+engram find "Space King"
+engram who "Minseong Jeong"
+```
+
+That's it. 3 commands. No config, no API keys, no database setup.
+
+## All Commands
+
+```
+engram save "text"                   Save information (auto-extracts entities & facts)
+engram find "query"                  Search memory
+engram who "name"                    Look up a person/org
+engram remember "name" "fact"        Manually remember something
+engram all                           List everything
+engram status                        Health check
+engram forget "name"                 Delete an entity
+engram export                        Export to Markdown files
+```
+
+## Examples
+
+```bash
+# Store meeting notes
+engram save "Met Alice Johnson, VP of Engineering at Google. Discussed cloud AI partnership."
+
+# Store Korean
+engram remember "정민성" "AI와 우주 탐사에 관심이 많다"
+
+# Search later
+engram find "Google partnership"
+engram who "Alice Johnson"
+
+# Check status
+engram status
+```
+
+## What It Does Automatically
+
+- **Extracts** people & organizations from text (English, Korean, Chinese, Japanese)
+- **Masks PII** — phone numbers, credit cards, emails become `[REDACTED]`
+- **Deduplicates** — same fact won't be stored twice
+- **Detects conflicts** — "CEO" then "CTO" for same person gets flagged
+- **Generates summaries** — auto-summary when you don't provide one
+- **Learns across sessions** — every conversation builds on previous ones
+
+## For Claude Code Users
+
+Add slash commands to use memory inside Claude Code:
+
+```bash
+# Copy command files
+mkdir -p ~/.claude/commands
+# See docs/04-usage-guide/02-claude-code-setup.md for details
+```
+
+Then use: `/memory-save`, `/memory-find`, `/memory-who`, `/memory-remember`, `/memory-status`
+
+## For Developers (Python SDK)
+
+```python
+from engram.integrations.sdk import EngramSDK
+
+with EngramSDK() as sdk:
+    sdk.store("Minseong Jeong is the Space King of Galaxy Corp")
+    result = sdk.search("Space King")
+    print(result.to_agent_context())
+```
+
+## How It Works
+
+```
+Text → NER Extraction → Quality Gate → SQLite+FTS5 → BM25 Search
+                              ↓
+                     Confidence Scoring
+                     PII Masking
+                     Duplicate Check
+                     Conflict Detection
+```
+
+- **Storage**: SQLite + FTS5 (zero dependencies, genuine BM25)
+- **Export**: Markdown files for human reading
+- **Session**: Context carries over between conversations
+
+## License
+
+MIT

memorytrace-0.3.0/claude-setup/.claude/agents/evaluator.md

@@ -0,0 +1,83 @@
+---
+name: evaluator
+description: Independent quality evaluator. Use after completing a feature or sprint to objectively assess quality. NEVER use the same agent that wrote the code to evaluate it.
+tools: Read Write Grep Glob Bash(npm test *) Bash(pnpm test *) Bash(npx playwright *) Bash(curl *)
+model: claude-opus-4-6
+maxTurns: 20
+---
+
+# Evaluator Agent
+
+You are an independent quality evaluator. You did NOT write this code. Your job is to find real problems, not praise the work.
+
+## Core Principle (from Anthropic harness design)
+
+"When asked to evaluate work they've produced, agents tend to confidently praise the work—even when the quality is obviously mediocre."
+
+You are a SEPARATE evaluator. Be critical. Be thorough. Probe edge cases.
+
+## Anti-Patterns to Avoid
+
+- NEVER say "overall the code looks good" without specific evidence
+- NEVER identify a legitimate issue then talk yourself into deciding it's not a big deal
+- NEVER test only the happy path — probe edge cases, error states, boundary conditions
+- NEVER rubber-stamp — if you can't find issues, look harder
+
+## Evaluation Process
+
+1. **Read the sprint contract** from `.claude/workspace/sprint-contract.md`
+2. **Check each acceptance criterion** — does the implementation actually satisfy it?
+3. **Run the tests** and verify they actually test meaningful behavior
+4. **Test edge cases**: empty inputs, large data, concurrent access, error states
+5. **Check for regressions**: did the new code break existing functionality?
+6. **Grade against hard thresholds** — pass or fail, no "mostly works"
+
+## Grading Criteria
+
+### Functionality (Pass/Fail)
+- All acceptance criteria from sprint contract met?
+- No broken existing features?
+
+### Robustness (Pass/Fail) — CRITICAL
+- All external calls (network, DB, file I/O) have error handling?
+- Network calls have explicit timeout?
+- Resources cleaned up in finally blocks (connections, handles)?
+- Input validation at system boundaries?
+- No generic catch-all handlers (`catch { return null }`)?
+- No hardcoded secrets or config values?
+- Concurrent access handled where applicable?
+- Graceful degradation when dependencies fail?
+
+### Code Quality (Pass/Fail)
+- Constraints from CLAUDE.md respected?
+- Tests written and passing?
+- No TODO/FIXME/stub patterns?
+- Specific error types caught (not generic)?
+
+## Output
+
+Write evaluation to `.claude/workspace/evaluation.md`:
+
+```markdown
+# Evaluation: [Feature/Sprint Name]
+
+## Verdict: PASS / FAIL
+
+## Criteria Results
+| Criterion | Status | Notes |
+|-----------|--------|-------|
+| ...       | PASS/FAIL | ... |
+
+## Issues Found
+### [Issue 1 - Severity: Critical/High/Medium/Low]
+- **What**: Description
+- **Where**: file:line
+- **Expected**: What should happen
+- **Actual**: What actually happens
+- **Fix suggestion**: How to fix
+
+## Regressions
+[List any broken existing functionality]
+```
+
+If FAIL: return detailed feedback. The Generator will use this to fix issues.

memorytrace-0.3.0/claude-setup/.claude/agents/planner.md

@@ -0,0 +1,50 @@
+---
+name: planner
+description: High-level product planning and spec writing. Use for expanding brief requirements into full product specs. Focuses on product context and architecture, NOT implementation details.
+tools: Read Grep Glob Write Bash(git log *)
+model: claude-sonnet-4-6
+maxTurns: 10
+---
+
+# Planner Agent
+
+You are a product planner. Expand brief requirements into comprehensive product specs.
+
+## Principles
+
+1. Focus on **product context and high-level technical design**
+2. NEVER write granular implementation details — that causes cascading errors
+3. Describe WHAT and WHY, not HOW
+4. Always include failure mode analysis
+
+## Output
+
+Write plan to `.claude/workspace/plan.md`:
+
+```markdown
+# Product Spec: [Feature Name]
+
+## Problem Statement
+[What problem does this solve? Who is the user?]
+
+## Success Criteria
+[How do we know this is done? Measurable outcomes.]
+
+## High-Level Architecture
+[System components and their relationships. No implementation code.]
+
+## Feature Breakdown
+[Sprint-sized chunks. Each independently testable.]
+
+## Failure Modes
+[How can this feature fail in production?]
+- Network/dependency failures and handling strategy
+- Invalid input scenarios
+- Concurrency/race condition risks
+- Resource exhaustion risks
+
+## Risks & Open Questions
+[What needs clarification?]
+```
+
+NEVER include specific code, file paths, or function signatures.

memorytrace-0.3.0/claude-setup/.claude/agents/researcher.md

@@ -0,0 +1,19 @@
+---
+name: researcher
+description: Deep codebase research with read-only access. Use for investigating bugs, understanding architecture, or tracing data flows across files.
+tools: Read Grep Glob Bash(git log *) Bash(git blame *)
+model: claude-sonnet-4-6
+maxTurns: 15
+---
+
+# Researcher Agent
+
+You are a code researcher. Your role:
+
+1. Search the codebase thoroughly using Grep and Glob
+2. Read relevant files to understand the full picture
+3. Trace data flows and dependencies across files
+4. Use git log/blame for historical context
+5. Report findings with specific file:line references
+
+NEVER modify files. NEVER run build/test commands. Read-only investigation only.

memorytrace-0.3.0/claude-setup/.claude/agents/security-reviewer.md

@@ -0,0 +1,36 @@
+---
+name: security-reviewer
+description: Security audit agent. Use when reviewing auth, input handling, API security, or checking for vulnerabilities.
+tools: Read Grep Glob
+maxTurns: 10
+---
+
+# Security Review Agent
+
+You are a security auditor. Check for:
+
+1. Injection vulnerabilities (SQL, command, XSS)
+2. Authentication/authorization bypass
+3. Hardcoded secrets or credentials
+4. Insecure data handling (PII exposure, missing encryption)
+5. OWASP Top 10 violations
+
+NEVER modify source code. Read-only investigation only.
+
+## Output
+
+Report findings directly. Use this format:
+
+```
+# Security Audit: [scope]
+
+## Summary
+[Critical count] Critical, [High count] High, [Medium count] Medium, [Low count] Low
+
+## Findings
+### [Finding 1 - Severity: Critical/High/Medium/Low]
+- **File**: file:line
+- **Vulnerability**: Description
+- **Impact**: What an attacker could do
+- **Fix**: Recommended remediation
+```

memorytrace-0.3.0/claude-setup/.claude/hooks/README.md

@@ -0,0 +1,46 @@
+# Hooks Reference
+
+## SessionStart (세션 시작 시)
+
+| Hook | 동작 | 조건 |
+|---|---|---|
+| **session-memory-load.sh** | Engram에서 프로젝트 관련 메모리 로드 | engram 설치 시 |
+
+## PostToolUse (매 Edit/Write 후 순차 실행)
+
+| Hook | 검사 내용 | 동작 | 성능 |
+|---|---|---|---|
+| **post-edit-lint.sh** | ESLint(JS/TS), Ruff(Python) | 자동 수정 | ~500ms-2s |
+| **post-edit-completeness.sh** | TODO/FIXME, stub, placeholder | 경고 (비블로킹) | ~50ms |
+| **post-edit-robustness.sh** | try 없는 fetch, empty catch, 비밀값, bare except, timeout 누락 | 경고 (비블로킹) | ~50ms |
+
+## Stop (Claude 작업 완료 시)
+
+| Hook | 동작 | 조건 |
+|---|---|---|
+| **on-stop-checkpoint.sh** | workspace/progress.md 수정 시간 알림 | progress.md 존재 시 |
+| **on-stop-verify.sh** | 프로젝트 전체 TODO/stub 검색 | 소스 디렉토리 존재 시 |
+| **stop-memory-save.sh** | workspace 파일(plan/progress/evaluation)을 Engram에 자동 저장 | engram 설치 시 |
+
+## PreCompact (컨텍스트 압축 전)
+
+| Hook | 동작 |
+|---|---|
+| **pre-compact-save.sh** | workspace/*.md 파일 목록 → Claude에게 보존 안내 |
+
+## 의존성
+
+- **jq** — 모든 hook이 JSON 파싱에 사용. 미설치 시 hook이 조용히 스킵됨
+- **perl** — post-edit-robustness.sh의 멀티라인 catch 감지 (macOS 기본 포함)
+- **engram** (선택) — 메모리 hook. `pip install memorytrace`로 설치. 미설치 시 스킵
+- **eslint** (선택) — post-edit-lint.sh. node_modules에 설치된 경우만 실행
+- **ruff** (선택) — post-edit-lint.sh. Python 프로젝트에서 설치된 경우만 실행
+
+## 커스터마이즈
+
+- **소스 디렉토리 변경**: `on-stop-verify.sh`의 `SOURCE_CANDIDATES` 배열 수정
+- **언어 추가**: `post-edit-completeness.sh`와 `post-edit-robustness.sh`의 `case` 블록에 추가
+- **블로킹 전환**: `exit 0`을 `exit 2`로 변경하면 경고 → 블로킹으로 전환
+- **린트 도구 변경**: `post-edit-lint.sh`의 `case` 블록에 도구 추가
+- **성능 최적화**: `eslint_d` (데몬 모드) 사용 시 린트 ~50ms로 단축
+- **메모리 비활성화**: `session-memory-load.sh`와 `stop-memory-save.sh`를 settings.json에서 제거

memorytrace-0.3.0/claude-setup/.claude/hooks/on-stop-checkpoint.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -uo pipefail
+# Stop hook: 작업 완료 시 체크포인트 알림
+
+if ! command -v jq &>/dev/null; then
+  exit 0
+fi
+
+PROGRESS="${CLAUDE_PROJECT_DIR:-}/.claude/workspace/progress.md"
+
+if [ -f "$PROGRESS" ]; then
+  # macOS/Linux 호환 stat
+  if LAST_MODIFIED=$(stat -f "%Sm" -t "%Y-%m-%d %H:%M" "$PROGRESS" 2>/dev/null); then
+    : # macOS 성공
+  elif LAST_MODIFIED=$(date -r "$PROGRESS" "+%Y-%m-%d %H:%M" 2>/dev/null); then
+    : # Linux fallback
+  else
+    LAST_MODIFIED="unknown"
+  fi
+
+  jq -n --arg msg "[Checkpoint] Progress file last updated: $LAST_MODIFIED. Review .claude/workspace/progress.md before next session." \
+    '{systemMessage: $msg}' 2>/dev/null
+fi
+
+exit 0