memory-bank-skill 3.0.0__tar.gz

memory_bank_skill-3.0.0/.gitignore

@@ -0,0 +1,39 @@
+.installed-manifest.json
+*.backup.*
+*.pre-mb-backup.*
+.DS_Store
+.env
+*.env
+*.pem
+*.key
+*.secret
+__pycache__/
+*.pyc
+.coverage
+htmlcov/
+.pytest_cache/
+.ruff_cache/
+
+# Project-local memory bank (dogfood). Contains session-specific status, plans,
+# and progress logs that are not part of the distributed skill — they live in
+# /.memory-bank/ of every consumer project. Never ship them in the public skill repo.
+/.memory-bank/
+
+# Build artifacts
+/dist/
+/build/
+*.egg-info/
+
+# Cross-agent adapter output from dogfood / test installs in this repo
+# (the skill writes these into any project where `/install` runs — when we
+# dogfood, they end up at the repo root and shouldn't be tracked).
+/.cursor/
+/.windsurf/
+/.clinerules/
+/.kilocode/
+/.codex/
+/.opencode/
+/.mb-agents-owners.json
+/.mb-pi-manifest.json
+/opencode.json
+/AGENTS.md

memory_bank_skill-3.0.0/CHANGELOG.md

@@ -0,0 +1,278 @@
+# Changelog
+
+All notable changes to this project are documented here. The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and versioning follows [SemVer](https://semver.org/spec/v2.0.0.html).
+
+## [Unreleased]
+
+---
+
+## [3.0.0] — 2026-04-20
+
+First stable 3.x release. Combines the work from the `3.0.0-rc1` / `rc2` / `rc3`
+candidates: native cross-agent install (Claude Code, Codex, Cursor, OpenCode),
+five-artifact global parity for Cursor, byte-level idempotent `install.sh`,
+and the `memory-bank install` / `memory-bank uninstall` / `memory-bank doctor`
+CLI. Install via `pipx install memory-bank-skill` or `brew install fockus/tap/memory-bank`.
+
+### Fixed (idempotency, promoted from `3.0.0-rc3`)
+
+- **`install.sh` is now byte-level idempotent** — repeat runs on an up-to-date tree create **zero** `.pre-mb-backup.*` files.
+  - Root cause: `backup_if_exists()` did an unconditional `mv` on any existing target, and `install_file()` called it without comparing `src` / `dst` content. For localized files (`RULES.md`, Cursor paste-file) a raw `cmp -s src dst` never matched because `dst` had been language-substituted.
+  - Observed damage before the fix: 14 historic installs → 1628 accumulated `.pre-mb-backup.*` files; a single "clean" reinstall generated 48 backups, 37 of them byte-identical to the active file.
+- **New helpers** in `install.sh`:
+  - `install_file()` short-circuits via `cmp -s "$src" "$dst"` when content is already identical.
+  - `backup_if_exists()` accepts an optional 2nd arg `expected_content_path` and returns `2` (skip marker) on content match.
+  - `install_file_localized()` — composes expected post-install content in a temp file (`cp src` + `localize_path_inplace`), compares to `dst` via `cmp -s`, skips backup+write on match. Used for `RULES.md` in Step 1.
+  - `localize_path_inplace()` — substitution helper without the existence shortcut.
+  - `install_cursor_user_rules_paste()` rewritten with a compose-to-tmp + `cmp -s` skip instead of unconditional overwrite.
+- **Manifest `.backups[]`** is now filtered via `os.path.exists` — stale refs from user-cleaned `.pre-mb-backup.*` files are dropped instead of accumulating across installs.
+- **`tests/e2e/test_install_idempotent.bats`** — 5 bats scenarios: zero backups on repeat install, exactly-one backup per real content diff, zero backups after external delete, language-swap backs up only localize-target files, manifest lists only existing backup paths.
+
+**Result**: repeat install on an up-to-date tree → 0 backups. Language swap (`--language en` → `--language ru`) → exactly 2 backups (`RULES.md` + `memory-bank-user-rules.md`) and nothing else.
+
+### Added
+
+- **Cursor global parity — five artifacts under `~/.cursor/` are now installed unconditionally** (without `--clients cursor`):
+  - `~/.cursor/skills/memory-bank/` — symlink to the canonical skill bundle (Cursor auto-discovers personal skills).
+  - `~/.cursor/hooks.json` + `~/.cursor/hooks/*.sh` — three global hooks (`sessionEnd` / `preCompact` / `beforeShellExecution`); each entry is marked `_mb_owned: true`, while user hooks on the same events are preserved.
+  - `~/.cursor/commands/*.md` — user-level slash commands mirroring the skill `commands/` directory.
+  - `~/.cursor/AGENTS.md` — managed section with unique `memory-bank-cursor:start/end` markers, preserving user content above and below.
+  - `~/.cursor/memory-bank-user-rules.md` — paste-ready bundle for **Settings → Rules → User Rules** (Cursor has no file API for global User Rules; this is a one-time manual step per machine). The post-install hint prints `pbcopy`/`xclip` commands.
+- **`install.sh` helpers**: `cursor_agents_section()`, `install_cursor_global_agents()`, `install_cursor_user_rules_paste()`, `install_cursor_global_hooks()` (with jq-based `hooks.json` merge similar to Codex). `ensure_skill_aliases()` now creates the `~/.cursor/skills/memory-bank` alias.
+- **`uninstall.sh` Cursor branches**: preserve `~/.cursor/AGENTS.md` and `~/.cursor/hooks.json` from manifest-removal (managed merged files), remove only the memory-bank section and `_mb_owned` entries, delete `memory-bank-user-rules.md` and the `skills/memory-bank` alias, and `rmdir` empty `~/.cursor/{skills,hooks,commands}` directories.
+- **`tests/e2e/test_cursor_global.bats`** — 17 bats scenarios: install creates all five artifacts, idempotency, preserve-user-content in `AGENTS.md` + `hooks.json`, clean uninstall.
+- **`tests/pytest/test_cli.py::test_cli_install_uninstall_smoke_with_cursor_global`** — pytest smoke: real `install.sh` / `uninstall.sh` run in sandboxed `$HOME` with Cursor steps.
+- **Docs**:
+  - `SKILL.md` — Cursor promoted to the "native full support" tier, plus a new **Host-specific notes → Cursor** subsection with the five-artifact table and paste flow.
+  - `docs/cross-agent-setup.md` — supported clients table, **Cursor (full global parity + project adapter)** section, resource availability matrix with a Cursor column and a "Global rules" row, troubleshooting Q&A for User Rules.
+  - `README.md` — global install hint updated (Cursor included in baseline `memory-bank install`), new "Cursor-only quick start" section, adapter table extended with global artifacts.
+
+### Fixed
+
+- **`adapters/cursor.sh` — removed duplicate `# Global Rules` heading** in `.cursor/rules/memory-bank.mdc`. Previously the script wrote its own `# Global Rules` and then concatenated `rules/RULES.md`, which starts with the same heading, so the MDC file ended up with two identical H1s.
+
+### Changed
+
+- `install.sh` / `uninstall.sh` — added constants `CURSOR_DIR`, `CURSOR_SKILL_ALIAS`, `CURSOR_USER_RULES_FILE`, `CURSOR_START_MARKER` / `CURSOR_END_MARKER` (independent from Codex markers — allowing coexistence without conflicts in `~/.codex/AGENTS.md` vs `~/.cursor/AGENTS.md`).
+
+### Added (rolled up from prior Unreleased)
+
+- **Interactive client picker in `install.sh`** — when `--clients` is not set and stdin is a TTY, a multi-select menu is shown for 8 clients (`claude-code`, `cursor`, `windsurf`, `cline`, `kilo`, `opencode`, `pi`, `codex`). Accepts numbers, names, `all`, or empty input (`claude-code`). Suppressed by `--non-interactive` or non-TTY stdin.
+- **`install.sh --non-interactive`** — explicit interactive bypass for CI / scripted installs.
+- **Env `MB_CLIENTS`** — alternate way to set clients (same semantics as `--clients`), useful in Docker / pipx wrappers.
+- **`memory-bank install --non-interactive`** — forwards the flag from the Python CLI.
+- **`/mb install` subcommand** (section in `commands/mb.md`) — installs adapters from inside Claude Code / OpenCode / Codex. Uses `AskUserQuestion` for multi-select in CC and inline prompt in other agents. The `/mb` namespace protects against collisions with other skills.
+- **Windows compromise — Git Bash / WSL support**:
+  - `cli.py` no longer hard-fails on Windows. Added `find_bash()` with priority order: `MB_BASH` env override → `bash.exe` on PATH → `C:\Program Files\Git\bin\bash.exe` → WSL fallback.
+  - `system32\bash.exe` (WSL launcher shim) is ignored in favor of Git Bash / explicit WSL.
+  - `memory-bank doctor` now prints the resolved bash path on any platform.
+  - Missing bash on Windows now yields an actionable install hint (`winget` / WSL).
+- **README: full command reference** — two tables (18 top-level slash commands + 20 `/mb` subcommands), replacing the previous partial list of 23 lines.
+- **README: 3 ways to install cross-agent adapters** — interactive menu, CLI flags, `/mb install` from inside the agent.
+
+### Changed (rolled up from prior Unreleased)
+
+- `memory-bank install / uninstall / init / doctor` — removed `require_posix()` calls; now work on Windows when bash is available.
+- `tests/pytest/test_cli.py` — updated for the new platform model (29 tests, including 9 new `find_bash()` discovery tests + WSL wrapper mode).
+- `tests/bats/test_install_interactive.bats` — new file with 13 tests for CLI flags, validation, env overrides.
+
+### Docs (rolled up from prior Unreleased)
+
+- README: Platform matrix expanded (macOS native / Linux native / Windows Git Bash / Windows WSL / Windows without bash — the last one with a hint).
+- README: command tables numbered as "18 top-level + 20 `/mb` subcommands".
+
+---
+
+## [3.0.0-rc1] — 2026-04-20
+
+### Repository moved
+
+This skill moved to a new public repository: **[`fockus/skill-memory-bank`](https://github.com/fockus/skill-memory-bank)**. The old name `claude-skill-memory-bank` described only one of the 8 supported clients and had become misleading.
+
+**Migration guide:** [docs/repo-migration.md](docs/repo-migration.md).
+
+Historical releases (`v1.x`, `v2.0.0`, `v2.1.0`, `v2.2.0`) remain available in the old repository. `v3.0.0` and later are published in the new one.
+
+### Added
+
+- **Stage 8 — Cross-agent adapters** (7 clients beyond Claude Code): Cursor (CC-compat hooks), Windsurf (Cascade), Cline (`.clinerules/hooks/`), Kilo (+ git-hooks fallback), OpenCode (TS plugin with `experimental.session.compacting`), Codex (experimental hooks), Pi Code (dual-mode). See `docs/cross-agent-setup.md`.
+- **`adapters/_lib_agents_md.sh`** — refcount-based shared `AGENTS.md` ownership via `.mb-agents-owners.json`. Enables safe coexistence of OpenCode / Codex / Pi installs.
+- **`adapters/git-hooks-fallback.sh`** — universal `.git/hooks/` installer (post-commit auto-capture, pre-commit `<private>` warnings, chain pattern preserving user hooks).
+- **`install.sh --clients <list>`** — non-interactive multi-client install. Valid: `claude-code, cursor, windsurf, cline, kilo, opencode, pi, codex`.
+- **`docs/cross-agent-setup.md`** — complete per-client cheatsheet + hook capability matrix + troubleshooting FAQ.
+- **`docs/repo-migration.md`** — upgrade instructions for existing users.
+
+### Changed
+
+- Repository name: `claude-skill-memory-bank` → `skill-memory-bank`.
+- Install target directory: `~/.claude/skills/skill-memory-bank/` (previously `claude-skill-memory-bank/`).
+- `mb-upgrade.sh` now tracks the new origin URL.
+
+### Tests
+
+- **340+/340+ bats + e2e green** (+90 adapter bats, +10 install-clients e2e over v2.2.0).
+
+---
+
+## [2.2.0] — 2026-04-20
+
+Knowledge-reach release: cold-start through JSONL import, code graph for 6 languages, tags normalization through a controlled vocabulary.
+
+### Added
+
+- **`/mb import --project <path> [--since] [--apply]`** (`scripts/mb-import.py`) — bootstrap Memory Bank from Claude Code transcripts (`~/.claude/projects/<slug>/*.jsonl`). Extracts daily-grouped `progress.md` sections and arch-discussion notes (≥3 consecutive assistant messages >1K chars). Dedup via SHA256(timestamp + first 500 chars), resume through `.import-state.json`. PII auto-wrap (email + API-key → `<private>`) integrated with v2.1 Stage 3. `--dry-run` by default.
+- **`/mb graph [--apply] [src_root]`** (`scripts/mb-codegraph.py`) — code graph for Python (stdlib `ast`, always-on) + Go / JavaScript / TypeScript / Rust / Java (through tree-sitter, opt-in). Nodes = module/function/class, edges = import/call/inherit. Output: `codebase/graph.json` (JSON Lines, grep/jq friendly) + `codebase/god-nodes.md` (top-20 by degree). SHA256 incremental cache per file. `HAS_TREE_SITTER` flag provides graceful degradation without grammars. Skipped dirs: `.venv`, `node_modules`, `__pycache__`, `.git`, `target`, `dist`, `build`.
+- **`/mb tags [--apply] [--auto-merge]`** (`scripts/mb-tags-normalize.sh`) — Levenshtein-based synonym detection + merge through a controlled vocabulary. Distance ≤2 detection, ≤1 for `--auto-merge`. Vocabulary in `.memory-bank/tags-vocabulary.md` (user-editable) → fallback to `references/tags-vocabulary.md` (35 default tags). Exit 2 when unknown tags are found (drift signal).
+- **`references/tags-vocabulary.md`** — template with 35 default tags (Core: arch/auth/bug/perf/refactor/test/...; Process: debug/review/post-mortem/adr/spike; Workflow: blocked/todo/wip/imported/discussion).
+- **`mb-index-json.py`** — auto-kebab-case for tags: `FooBar → foo-bar`, `AUTH → auth`, `someThing → some-thing`. Dedup while preserving order. Source files remain untouched — only the index changes.
+
+### Changed
+
+- **`mb-codegraph.py`** — the Python-only v1 path was extended with a tree-sitter adapter for 5 new languages. Same node/edge schema. Lazy parser loading per language through `_TS_PARSERS` cache.
+- **`commands/mb.md`** — `/mb import`, `/mb graph`, `/mb tags` sections with full examples, dogfood outputs, and limitations.
+- **`install.sh`** — now prints a hint for tree-sitter extras.
+
+### Tests
+
+- pytest: **96** (was 44 after v2.1) — +17 `test_import`, +21 `test_codegraph` (Python), +14 `test_codegraph_ts` (tree-sitter).
+- bats: **208** (was 194) — +14 `test_tags_normalize`.
+- shellcheck: 0 warnings. ruff: all passed.
+
+### Gate v2.2 — passed ✅
+
+1. ✅ `/mb import` on real JSONL (2573 events) — **0.127s** (target ≤30s).
+2. ✅ `mb-codegraph.py` on `scripts/` — **0.068s** for 60 nodes + 487 edges (target ≤30s).
+3. ✅ Tags normalization: `sqlite_vec → sqlite-vec` auto-merged at distance=1, 2 notes rewritten. Unknown tag → exit 2 drift signal works.
+4. ✅ Full regression: 96 pytest + 208 bats + ruff + shellcheck clean.
+5. ✅ VERSION 2.2.0, CHANGELOG updated.
+
+### ADR pivots
+
+- **ADR-006 updated (Stage 6.5):** tree-sitter originally started as deferred opt-in. After user feedback ("I often work with Node/Go") it was implemented in v2.2, with graceful degradation when grammars are absent. The Python path remained zero-dependency. See BACKLOG (Stage 6.5 shipped entry).
+
+### Deferred to v3.x backlog
+
+- Haiku-powered compression for `/mb import` summaries (currently deterministic first+last chars).
+- Debug-session detection in `/mb import` for `lessons.md`.
+- Type inference in `/mb graph` (edges are currently name-based and do not distinguish modules with same-named functions).
+- tree-sitter grammars for C/C++/Ruby/PHP/Kotlin/Swift (add on demand via `_TS_LANG_CONFIG`).
+
+## [2.1.0] — 2026-04-20
+
+Hardening release: auto-capture when `/mb done` is forgotten, drift detection without AI, PII protection in notes, status-based decay for old plans and notes.
+
+### Added
+
+- **Auto-capture SessionEnd hook** (`hooks/session-end-autosave.sh`) — if a session closes without `/mb done`, the hook appends a placeholder entry to `progress.md`. Lock file `.memory-bank/.session-lock` is written by `/mb done` → the hook sees a fresh lock (<1h) and skips auto-capture. `MB_AUTO_CAPTURE` env: `auto` (default) / `strict` / `off`. Concurrent-safe through `.auto-lock` (30s TTL). Idempotent per `session_id`.
+- **Drift checkers without AI** (`scripts/mb-drift.sh`) — 8 deterministic checkers (path, staleness, script_coverage, dependency, cross_file, index_sync, command, frontmatter). Output: `drift_check_<name>=ok|warn|skip` + `drift_warnings=N`. Exit 0 when 0 warnings, otherwise 1. `agents/mb-doctor.md` Step 0 = `mb-drift.sh` → LLM call only if `drift_warnings > 0`. Saves AI tokens.
+- **PII markers `<private>...</private>`** in notes — block contents do not enter `index.json` (summary + tags filtered), and `mb-search` replaces them with `[REDACTED]` (inline) or `[REDACTED match in private block]` (multi-line). Unclosed `<private>` is fail-safe: tail to EOF is treated as private. Entries get `has_private: bool`. `--show-private` requires `MB_SHOW_PRIVATE=1` env (double confirmation). `hooks/file-change-log.sh` warns when a `.md` file with a `<private>` block is Write/Edit-ed.
+- **Compaction decay `/mb compact`** (`scripts/mb-compact.sh`) — status-based archival: requires **age > threshold AND done-signal**. Active plans (not done) are **NOT archived** even if >180d — warning only. Done-signal (OR): file physically under `plans/done/`, OR a `✅`/`[x]` marker in `checklist.md`, OR mentioned as "done|closed|shipped" in `progress.md`/`STATUS.md`. Notes — `importance: low` + mtime >90d + no references in core files. `--dry-run` (default) reasons only; `--apply` executes and touches `.last-compact`. Entries under `notes/archive/` get `archived: bool`. `mb-search --include-archived` is opt-in for archive search.
+
+### Changed
+
+- **`agents/mb-doctor.md`** — Step 0 is now `mb-drift.sh`; LLM steps (1-4) run only when `drift_warnings > 0` or `doctor-full`.
+- **`scripts/mb-index-json.py`** — parser for `<private>` blocks, `has_private: bool` + `archived: bool` fields in the entry schema.
+- **`scripts/mb-search.sh`** — span-aware Python filter for REDACTED replacement. Added `--show-private` + `--include-archived`.
+- **`settings/hooks.json`** — SessionEnd event added for auto-capture.
+- **`commands/mb.md`** — `/mb done` writes `.session-lock`. Added `/mb compact` section with full status-based logic and examples.
+- **`SKILL.md`** — "Private content" and "Auto-capture" sections.
+- **`references/metadata.md`** — schema extended with `has_private` + `archived` fields.
+
+### Tests
+
+- bats: **194** (20 `test_compact` + 4 `test_search_archived` + 5 `test_search_private` + 20 `test_drift` + 12 `test_auto_capture` + regressions).
+- pytest: **44** (7 PII + 2 archived + regressions).
+- e2e: 18 install/uninstall (including SessionEnd hook roundtrip).
+- shellcheck: 0 warnings (SC1091 info expected for `source _lib.sh`).
+- ruff: all passed.
+
+### Gate v2.1 — passed ✅
+
+1. ✅ Auto-capture end-to-end: simulated SessionEnd → `progress.md` updated without `/mb done`.
+2. ✅ `mb-drift.sh` on broken fixture: 7 warnings across 8 categories (≥5 target).
+3. ✅ PII security smoke: `TOP-SECRET-LEAK-CHECK-GATE21` inside `<private>` → **0 matches** in `index.json`.
+4. ✅ `/mb compact` dogfood: live bank is clean (0 candidates), artificial 150d done-plan → archived, 150d active-plan → not archived (safety works).
+5. ✅ CI matrix `[macos, ubuntu]` × (bats + e2e + pytest) green.
+
+### Deferred to backlog
+
+- LLM upgrade for auto-capture (currently append-only; details are reloaded by `/mb start` from JSONL).
+- `/mb done` weekly prompt for compaction checks.
+- Pre-commit drift hook as a separate file (YAGNI; documented in `references/templates.md`).
+
+## [2.0.0] — 2026-04-19
+
+Large refactor: the skill becomes language-agnostic, tested, CI-covered, and integrated with the Claude Code ecosystem. Three concepts under one roof: **Memory Bank + RULES + dev toolkit**.
+
+### Added
+
+- **Language detection (12 stacks)**: Python, Go, Rust, Node/TypeScript, Java, Kotlin, Swift, C/C++, Ruby, PHP, C#, Elixir. `scripts/mb-metrics.sh` emits key=value metrics for any of them.
+- **Override for project metrics**: `.memory-bank/metrics.sh` (priority 1 over auto-detect).
+- **`scripts/_lib.sh`** — shared utilities (workspace resolver, slug, collision-safe filename, detect_stack/test/lint/src_glob). 7 functions, 36 bats tests.
+- **`scripts/mb-plan-sync.sh`** and **`scripts/mb-plan-done.sh`** — automate plan↔checklist↔plan.md consistency through `<!-- mb-stage:N -->` markers.
+- **`scripts/mb-upgrade.sh`** — `/mb upgrade` self-update for the skill from GitHub (git fetch → prompt → ff-only pull + reinstall).
+- **`scripts/mb-index-json.py`** — pragmatic index for `notes/` (frontmatter) + `lessons.md` (H3 markers). Atomic write. PyYAML opt-in with fallback.
+- **`mb-search --tag <tag>`** — tag filtering via `index.json`.
+- **`/mb init [--minimal|--full]`** — unified initialization. `--full` (default) = `.memory-bank/` + `RULES.md` copy + stack auto-detect + `CLAUDE.md` + optional `.planning/` symlink.
+- **`/mb map [focus]`** — scan the codebase and generate `.memory-bank/codebase/{STACK,ARCHITECTURE,CONVENTIONS,CONCERNS}.md`.
+- **`/mb context --deep`** — full codebase content (default = 1-line summary).
+- **Frontend rules (FSD)** and **Mobile (iOS + Android)** in `rules/RULES.md` and `rules/CLAUDE-GLOBAL.md`.
+- **`MB_ALLOW_NO_VERIFY=1`** — bypass for `--no-verify` in `block-dangerous.sh`.
+- **Log rotation** for `file-change-log.sh`: >10 MB → `.log.1 → .log.2 → .log.3`.
+- **GitHub Actions** `.github/workflows/test.yml`: matrix `[ubuntu-latest, macos-latest]` × (bats + e2e + pytest) + lint job (shellcheck + ruff).
+- **148 bats tests** (unit + e2e + hooks + search-tag), **35 pytest tests**, **94% total coverage**.
+
+### Changed
+
+- **`codebase-mapper` → `mb-codebase-mapper`** (MB-native): output path `.planning/codebase/` → `.memory-bank/codebase/`; 770 lines → 316 (−59%); 6 templates → 4 (STACK, ARCHITECTURE, CONVENTIONS, CONCERNS), each ≤70 lines.
+- **`/mb update` and `mb-doctor`** no longer hardcode `pytest`/`ruff`/`src/taskloom/` — they use `mb-metrics.sh`.
+- **`SKILL.md` frontmatter** — invalid `user-invocable: false` replaced by `name: memory-bank` with a three-in-one concept description.
+- **`Task(...)` → `Agent(subagent_type=...)`** in all skill files. Grep check: 0 occurrences of `Task(`.
+- **`mb-doctor`** fixes desynchronization primarily through `mb-plan-sync.sh`/`mb-plan-done.sh`; Edit is reserved for semantic issues.
+- **`install.sh`** — always writes the `[MEMORY-BANK-SKILL]` marker when creating a new `CLAUDE.md` (previously only when merging into an existing one).
+- **`mb-manager actualize`** — now calls `mb-index-json.py` instead of manually writing `index.json`.
+- **`file-change-log.sh`** — removed `pass\s*$` from the placeholder regex (false positive); placeholder search now ignores Python docstrings.
+- **`install.sh` banner**: `19 dev commands` → `18 dev commands` (after init-command consolidation).
+
+### Deprecated
+
+- (none — all deprecations in this release are full removals; see Removed)
+
+### Removed
+
+- **`/mb:setup-project`** — merged into `/mb init --full`. `commands/setup-project.md` removed.
+- **Orphan agent `codebase-mapper`** — replaced with `mb-codebase-mapper`.
+- **Hardcoded `pytest -q` / `ruff check src/`** in `commands/mb.md` and `agents/mb-doctor.md`.
+- **All `Task(...)` invocations** in skill files (0 left).
+
+### Fixed
+
+- **E2E-found bug #1**: `install.sh` did not add the `[MEMORY-BANK-SKILL]` marker when creating a **new** `CLAUDE.md`. Result: `uninstall.sh` could not find the section to clean.
+- **E2E-found bug #2**: `uninstall.sh` used the GNU-only `realpath -m` flag. BSD `realpath` on macOS failed. Fix: the manifest already stores absolute paths, so `realpath` is unnecessary.
+- **Node `src_glob`**: brace-pattern `*.{ts,tsx,js,jsx}` replaced by space-separated globs for portable grep.
+- **`mb-note.sh`**: name collision (two notes in one minute) now yields `_2/_3` suffixes (previously `exit 1`).
+- **`file-change-log` false positives**: bare `pass` in Python, TODO inside docstrings.
+- **shellcheck SC1003** in the awk hook block — rewritten through `index()` without nested single-quote escapes.
+
+### Security
+
+- **`block-dangerous.sh`** updated with `MB_ALLOW_NO_VERIFY=1` explicit opt-in override — previously `--no-verify` was blocked with no safe escape.
+- **Secrets detection** in `file-change-log.sh` continues to work (`password|secret|api_key|token|private_key` in source code).
+
+### Infrastructure
+
+- **Dogfooding**: the skill now uses `.memory-bank/` in its own repository. The v2 refactor plan lives in `.memory-bank/plans/`, and sessions are closed through `/mb done`.
+- **VERSION marker**: `2.0.0-dev` → `2.0.0` written by `install.sh` into `~/.claude/skills/memory-bank/VERSION`.
+- **CI green** on macOS + Ubuntu, 0 shellcheck warnings, ruff all passed.
+
+---
+
+## [1.0.0] — 2025-10-XX (pre-refactor baseline)
+
+- Initial Memory Bank skill: `.memory-bank/` structure, `/mb` roadmap command, 4 agents, 2 hooks, 19 commands.
+- Python-first: hardcoded `pytest`, `ruff`, `src/taskloom/`.
+- Orphan artifacts from GSD: `codebase-mapper`, `.planning/`.
+- 0 automated tests.
+
+[2.0.0]: https://github.com/fockus/claude-skill-memory-bank/releases/tag/v2.0.0
+[1.0.0]: https://github.com/fockus/claude-skill-memory-bank/releases/tag/v1.0.0

memory_bank_skill-3.0.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Anton Ivanov
+
+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.