Przejdź do treści
Biblioteka skilli
AI i agenci · Analiza

claude-memory-graph

/claude-memory-graph

Buduje interaktywny graf pamięci Claude Code z jednego pliku HTML: powiązania, filtry typów i szczegóły każdej notatki.

Licencja
MIT
Ostatnia zmiana repozytorium
10.11.2026

Ocena i bezpieczeństwo

100/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
7
Skrypty
2
Komentarze HTML
0
Adresy URL w SKILL.md
0
Przestarzałość

Brak sygnałów. Repozytorium zmienione 10.11.2026, bez odwołań do wycofanych modeli i API.

Licencja

MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.

Co sprawdzić przed instalacją
  • 2 skrypty w paczce. Przejrzyj je przed instalacją, bo Claude może je uruchomić.

Skrypty: build.py, test_build.py

Co robi i kiedy

Buduje interaktywny graf pamięci Claude Code z jednego pliku HTML: powiązania, filtry typów i szczegóły każdej notatki.

Oryginalny opis (EN)

Visualize Claude Code's memories as an interactive graph. Builds one self-contained HTML file from every memory in ~/.claude/projects/*/memory, grouped by workspace, with [[links]] as edges, search, type filters and a detail panel. Use when the user asks to see, visualize, map or graph their Claude memories, wants to know what Claude remembers about them or their projects, or wants to refresh the graph after editing or cleaning up memories.

Jak działa

  1. Build
  2. How it reads memories
  3. Using the page
  4. Rules
Sekcje z nagłówków SKILL.md, w oryginale · 55 linii, 781 słów
Podgląd SKILL.md
---
name: claude-memory-graph
description: Visualize Claude Code's memories as an interactive graph. Builds one self-contained HTML file from every memory in ~/.claude/projects/*/memory, grouped by workspace, with [[links]] as edges, search, type filters and a detail panel. Use when the user asks to see, visualize, map or graph their Claude memories, wants to know what Claude remembers about them or their projects, or wants to refresh the graph after editing or cleaning up memories.
---

# Claude memory graph

The skill draws an interactive map of everything Claude Code remembers, across every project. Each memory is a node. Memories cluster around their project, and projects sit inside workspace regions. `[[links]]` between memories become edges.

## Build

Run the builder that sits next to this file (Python 3.8+, stdlib only):

```bash
python3 -I <this skill's directory>/build.py <output.html>
```

For a typical install that is:

```bash
python3 -I ~/.claude/skills/claude-memory-graph/build.py ~/claude-memory-graph.html
```

- If the user named an output path, use theirs. Otherwise pick a place they can easily open, and tell them the full path.
- If the user plans to share the graph, add `--no-bodies` so only names and descriptions are embedded, not the full memory text.
- Options: `--root` points at a different projects dir (by default `$CLAUDE_CONFIG_DIR/projects` or `~/.claude/projects`). `--home` sets the home path used to shorten labels.
- The builder prints `N memories, N links, N dangling, N projects, N workspaces -> path`. Report those numbers.
- It asserts that node ids are unique and that every edge points at a real node. If an assertion fails, report the error. Don't hand over a broken file.

On Windows, the interpreter is usually `python` or `py` rather than `python3`.

Open the file in a browser for the user if your environment allows it (`open` on macOS, `xdg-open` on Linux, `start ""` on Windows). If a sandbox blocks that, give them the command to run.

## How it reads memories

- **Nodes:** every `*.md` in each `<projects>/<dir>/memory/` except `MEMORY.md`. Temporary session dirs (`-private-…`, `-tmp-…`, `-var-folders-…`) are skipped.
- **Fields:** frontmatter `name`, `description` and `type`, whether top-level or nested under `metadata:`. A missing `type` shows as `unknown`.
- **Edges:** `[[target]]` in a memory's body. A link resolves by `name`, then by filename, and prefers the same project. Links that don't resolve are counted as "dangling".
- **Projects:** a dir's name is its working directory with every non-alphanumeric character replaced by `-`. The builder strips the home prefix and one container folder (`Projects`, `code`, `src`, `repos`, …), so `-Users-ana-Projects-acme-api` becomes `acme-api`. The home dir itself is `home`.
- **Workspaces:** a project joins the workspace of the shortest other project whose label prefixes it at a `-`, so `acme-api` and `acme-web` both join `acme`. Projects with no parent dir form their own workspace.

## Using the page

- Hover a node for its name and date. Click one for its description, file path and links in both directions.
- Search opens a results list in the side panel, grouped by workspace and showing a snippet of each description. Enter opens the first result. Clicking a result centers it and shows its details, with "← results" to go back.
- Clicking a memory opens a reader with its full text. `[[links]]` inside it jump to the linked memory.
- The theme menu offers Auto (follows the system light/dark setting), Paper, Ink, Blueprint, Neon, Print, Claude, Cursor, Skeuo and Cyber. Zoom with the +/−/fit buttons, F to fit, / to search. The choice is remembered in that browser.
- The type chips toggle memory types.
- Legend: clicking a workspace, project or type focuses it (shows only it), and clicking it again shows everything. ⇧/⌘-click adds or removes items; hovering a row previews it on the canvas; "Show all" resets the filters.
- Drag to pan, scroll to zoom, and drag nodes to move them. Esc closes the panel.

## Rules

- The page is read-only and never changes memories.
- The HTML embeds every memory's name, description and full text, unless it was built with `--no-bodies`. Tell the user that before they share it, and never commit it to a public repo.
- The output is a single offline file with no network requests. The one exception is Omarchy, where the page also loads its local `<output>.theme.js` sidecar (see below). Keep it that way if you change the template.
- On Omarchy, the builder reads the active theme from `~/.local/state/omarchy/current/theme/` (or the older `~/.config/omarchy/current/theme/`) and the page defaults to a matching "Omarchy" theme. The builder also writes `<output>.theme.js` beside the graph; the open page re-reads it every 2 seconds and re-themes itself live. To follow theme switches without a rebuild, install the bundled hook once with `omarchy hook install theme-set <this skill's directory>/hooks/omarchy-theme-set`: it runs `build.py --theme-only <output.html>` on every Omarchy theme change, which refreshes only that file. The hook assumes the graph is at `~/claude-memory-graph.html`; set `CLAUDE_MEMORY_GRAPH` in it for another path. Keep the two files together if the graph is moved.
- To add a theme, add a `:root[data-theme="…"]` token block to the template's CSS and an entry to its `THEMES` array. The canvas reads the same tokens.
Pliki w paczce (7)
  • .gitignore 1 KB
  • LICENSE 1 KB
  • README.md 5 KB
  • SKILL.md 5 KB
  • build.py 12 KB
  • template.html 48 KB
  • test_build.py 5 KB

Wersje

W bazie nie ma innych skilli o nazwie claude-memory-graph.

Migawka bazy z 13.11.2026 · wersja paczki 9fb9a3ce8d06 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 100audit-memories
    Przegląda zapisane pamięci agenta AI: inwentaryzuje je według typu i wieku, wskazuje nieaktualne, sprzeczne lub osierocone i prowadzi przez decyzje o zachowaniu, scaleniu albo usunięciu.
  • 90clean-claude-memory
    Audytuje, czyści, przepisuje lub przenosi pamięć projektu Claude Code, usuwa nieaktualne wpisy i odbudowuje MEMORY.md.
  • 100memory
    Zarządza trwałą warstwą pamięci w .claude/memory/: wyświetla, przeszukuje, zapisuje, aktualizuje, usuwa i deduplikuje wpisy.
  • 100memory-tool
    Daje agentom Claude trwałą pamięć między rozmowami przez narzędzie pamięci API, czyli plikowy magazyn pod /memories obsługiwany przez aplikację.
  • 98knowledge
    Przegląda, przeszukuje i porządkuje pamięć agentów między sesjami oraz wykrywa nieaktualne wpisy i katalogi.
  • 84stateless
    Sprawdza i wyłącza automatyczną pamięć Claude Code, czyli notatki zapisywane per repozytorium w ~/.claude/projects/PROJECT/memory/, z akcjami status, disable i czyszczeniem pamięci.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.