← ClaudeAtlas

archive-indexlisted

Build a hierarchical read-only index over a folder of old notes, drafts or exports so an agent can locate files without loading them all into context. Use when a directory holds hundreds of files and the agent keeps grepping blindly or asking "where is X".
xiaozhanglbz/astrolabe-index · ★ 0 · AI & Automation · score 72
Install: claude install-skill xiaozhanglbz/astrolabe-index
# Archive Index (Astrolabe) Turns a folder into a three-level catalogue an agent can navigate: **volume → chapter → page**. Nothing in the folder is modified; the index is generated data that can be deleted and rebuilt at any time. ## When to use this Use it when all of these hold: - The folder holds many files (roughly 50+), accumulated over time. - Files come in families: dated copies, versions, exports, per-project folders. - You keep needing to locate a specific file, or answer "what is in here". - Loading the whole tree into context is too expensive. ## Steps 1. **Build the index** (writes `<dir>/_index/`, touches nothing else): ```bash python3 -m astrolabe build <dir> ``` For a flat folder with very large subfolders, lower the split threshold so chapters stay navigable: ```bash python3 -m astrolabe build <dir> --split-threshold 50 ``` 2. **Read `_index/00_catalog.json` first.** It is a few KB. It contains `volume_list` with per-volume `pages`, `chapters`, `keywords`, `heat` (`hot` < 30 days, `warm` < 180, `cold` older) and `entry` paths. 3. **Pick a volume, then read its file** (`_index/NN_<volume>.json`). `chapter_list` gives one `entry` path per chapter, plus `keywords` and either `page_list` (small chapter) or `sample_pages` + `page_distribution` (large chapter — list the directory to see everything). 4. **Open pages by entry path.** Entries are relative to the indexed root and always end with `/` for directories.