[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M756X3EYD0GQM09MZVMVR312":3,"skill-same:01M756X3EYD0GQM09MZVMVR312":46},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":14,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":29,"ownerDescription":30,"updatedAt":31,"install":32,"resources":45},"01M756X3EYD0GQM09MZVMVR312","leann-index",null,"Build, update, rebuild, watch, or list LEANN indexes. Use for `leann build`, `build-jsonl`, `rebuild`, `watch`, source readers, backend choice, incremental behavior, metadata stamping, folder filters, embedding cache, and AST-aware code indexing.","zm2231\u002FLEANN-fork","zm2231","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F25645999?v=4","01M6QV5TB1EMPJXZPWJ8FX9PR7","Zain Merchant","01M756WNJF91D9QKB1FX72KRE9",0,"MIT","https:\u002F\u002Fgithub.com\u002Fzm2231\u002FLEANN-fork\u002Fblob\u002Fzain-canonical\u002Fskills\u002Fleann-index\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: leann-index\ndescription: Build, update, rebuild, watch, or list LEANN indexes. Use for `leann build`, `build-jsonl`, `rebuild`, `watch`, source readers, backend choice, incremental behavior, metadata stamping, folder filters, embedding cache, and AST-aware code indexing.\n---\n\n# LEANN Indexing\n\nUse this when creating or maintaining indexes under `.leann\u002Findexes\u002F\u003Cname>\u002F`.\n\n## First Decision: Backend\n\nChoose by corpus size and whether content is edited in place.\n\n| Backend | Best use | Incremental `--docs` | Incremental `build-jsonl --incremental-by-id` |\n|---|---|---|---|\n| `flat` | Small\u002Fmedium edited corpora, exact recall | Add\u002Fremove\u002Fmodify in place | Add\u002Fremove\u002Fmodify in place |\n| `ivf` | Larger edited corpora, approximate FAISS search | Add\u002Fremove\u002Fmodify in place | Add\u002Fremove\u002Fmodify in place |\n| `hnsw` | Large mostly static corpus that fits in RAM | Any docs change triggers full rebuild | Add-only for non-compact indexes; changed\u002Fremoved rows rebuild |\n| `diskann` | Very large disk-backed corpus | Full rebuild | Full rebuild |\n| compact indexes | Archived\u002Fread-only indexes | No incremental | No incremental |\n\nDefault to `flat --no-recompute` for actively edited markdown\u002Fwiki\u002FCRM\u002Ftool-doc corpora. Default to `hnsw` for large mostly static corpora. Use `ivf` when flat is too slow but remove\u002Fmodify still matters.\n\n## Embeddings\n\nPick the embedding provider explicitly for reproducible builds. Examples:\n\n```bash\n--embedding-mode sentence-transformers \\\n--embedding-model facebook\u002Fcontriever\n```\n\nFor OpenAI-compatible embedding servers:\n\n```bash\n--embedding-mode openai \\\n--embedding-model \u003Cmodel-name> \\\n--embedding-api-base \u003Cbase-url> \\\n--embedding-api-key \u003Capi-key>\n```\n\nIf the server requires a non-empty placeholder key, pass one explicitly.\n\n## Build From Files\n\n```bash\nleann build docs --docs .\u002Fdocuments \\\n  --backend-name flat \\\n  --no-recompute \\\n  --embedding-mode sentence-transformers \\\n  --embedding-model facebook\u002Fcontriever\n```\n\n`build --docs` preserves file\u002Fpath metadata and stamps:\n\n- `indexed_at`\n- `created_at`\n- `modified_at`\n- `source_root`\n- `relative_path`\n- `folder_path`\n- `top_folder`\n\n### Folder Filters\n\nUse `--top-folder-depth` when the first path segment is only a wrapper folder.\n\n```bash\nleann build docs --docs \u002Fpath\u002Fto\u002Fdocs \\\n  --top-folder-depth 2 \\\n  --backend-name flat \\\n  --no-recompute \\\n  --embedding-mode sentence-transformers \\\n  --embedding-model facebook\u002Fcontriever\n```\n\nFor `Clients\u002FBD\u002Fproposal.md` with `--top-folder-depth 2`, metadata is:\n\n- `relative_path = Clients\u002FBD\u002Fproposal.md`\n- `folder_path = Clients\u002FBD`\n- `source_root = \u003Cdocs root>`\n- `top_folder = BD`\n\nThen search with:\n\n```bash\nleann search dropbox \"proposal\" \\\n  --metadata-filters '{\"top_folder\":{\"==\":\"BD\"}}'\n```\n\nDo not call this `category` in core LEANN. `top_folder` is factual filesystem metadata; app-specific category semantics belong in JSONL metadata or a source reader.\n\n## Build From JSONL\n\nUse `build-jsonl` when the caller already has one row per searchable object and needs exact metadata preserved.\n\n```jsonl\n{\"id\":\"workon\",\"text\":\"workon switch project repo context\",\"metadata\":{\"name\":\"workon\",\"group\":\"project\",\"tier\":\"hot\"}}\n```\n\n```bash\nleann build-jsonl code-mode-tools \\\n  --input .pi\u002Fcode-mode-leann\u002Ftools.jsonl \\\n  --text-field text \\\n  --metadata-field metadata \\\n  --id-field id \\\n  --incremental-by-id \\\n  --backend-name flat \\\n  --no-recompute \\\n  --embedding-mode sentence-transformers \\\n  --embedding-model facebook\u002Fcontriever\n```\n\nRules:\n\n- Every row needs a unique stable `--id-field`.\n- If `metadata.id` exists, it must match the top-level ID.\n- LEANN writes the stable ID into metadata as `id` and `source_document_id`.\n- `metadata.indexed_at` is ignored for rowhashes, so timestamp refreshes do not re-embed unchanged rows.\n\n## Rebuild And Watch\n\n```bash\nleann rebuild docs\nleann rebuild docs --force\nleann watch docs --interval 30\n```\n\n`rebuild` replays stored `build_config`, including docs paths, JSONL input, chunking fields, backend, embedding flags, `--top-folder-depth`, and `--incremental-by-id`.\n\nUse `--force` for a full rebuild. The content-addressed embedding cache keeps force rebuilds cheap when text is unchanged.\n\n## Code Indexes\n\n```bash\nleann build code --docs .\u002Fsrc --use-ast-chunking \\\n  --backend-name flat \\\n  --no-recompute \\\n  --embedding-mode sentence-transformers \\\n  --embedding-model facebook\u002Fcontriever\n```\n\nAST chunking supports Python, Java, C#, TS\u002FTSX\u002FJS. Other languages fall back to token chunking. Verify by reading `meta.json` after build.\n\n## Source Readers\n\nSpecialized readers emit richer source metadata, including `source_type`, `event_time`, `author`, and source-specific IDs when available.\n\n```bash\nleann index-browser chrome\nleann index-email\nleann index-calendar --max-count 1000\nleann index-imessage\nleann index-chatgpt --export-path ~\u002Fchatgpt-export.zip\nleann index-claude --export-path ~\u002Fclaude-export.zip\nleann index-wechat --export-dir ~\u002Fwechat-export\n```\n\nAll `index-*` commands share the same embedding flags as `leann build`.\n\n## Important Flags\n\nInput:\n\n- `--docs PATH [PATH ...]`\n- `--file-types .md,.pdf,.txt`\n- `--include-hidden` \u002F `--no-include-hidden`\n- `--top-folder-depth N`\n\nBackend:\n\n- `--backend-name {hnsw,diskann,ivf,flat}`\n- `--compact` \u002F `--no-compact`\n- `--recompute` \u002F `--no-recompute`\n- `--force`\n\nChunking:\n\n- `--doc-chunk-size N`, `--doc-chunk-overlap N`\n- `--code-chunk-size N`, `--code-chunk-overlap N`\n- `--use-ast-chunking`\n- `--ast-chunk-size N`, `--ast-chunk-overlap N`\n\nJSONL:\n\n- `--input FILE.jsonl`\n- `--text-field FIELD`\n- `--metadata-field FIELD`\n- `--id-field FIELD`\n- `--incremental-by-id`\n\nDefault `--docs` extensions include documents (`.txt`, `.md`, `.pdf`, `.docx`, `.pptx`), common code\u002Fconfig\u002Fmarkup files, and `.ipynb`. PDFs get PyMuPDF\u002Fpdfplumber first, then generic fallback.\n\n## Embedding Cache\n\nBuild-time embeddings use `~\u002F.leann\u002Fembed-cache.sqlite` by default.\n\n- Override: `LEANN_EMBED_CACHE_PATH=\u002Fpath\u002Fcache.sqlite`\n- Disable: `LEANN_EMBED_CACHE=0`\n- Force model revision key: `LEANN_EMBED_MODEL_REV=\u003Cstable-revision>`\n\nCache keys include canonicalization version, model, provider\u002Fmode, dimensions, normalization\u002Fpooling, prompt template, truncation settings, and model revision when known. Batch size and timeouts do not affect cache keys.\n\nBM25\u002FFTS5 is built and maintained by default. Do not pass a `--prebuild-bm25` flag in this fork.\n\n## Shipped: `--top-folder-depth`\n\nSHIP for the requested folder-filter change:\n\n- Added `leann build --top-folder-depth` in `packages\u002Fleann-core\u002Fsrc\u002Fleann\u002Fcli.py`.\n- Default is `1`, preserving current behavior.\n- With `--top-folder-depth 2`, `Clients\u002FBD\u002Fproposal.md` yields `top_folder = BD`.\n- Wired through build defaults, explicit option tracking, persisted `build_config`, rebuild, and watch replay.\n- Documented in `docs\u002Fconfiguration-guide.md` and `docs\u002FCHANGELOG.md`.\n- Verified with `uv run pytest tests\u002Ftest_cli_build_config.py tests\u002Ftest_indexed_at.py -q`: `22 passed`.\n- Verified with `uv run ruff check packages\u002Fleann-core\u002Fsrc\u002Fleann\u002Fcli.py tests\u002Ftest_cli_build_config.py tests\u002Ftest_indexed_at.py`.\n- Real CLI smoke confirmed `--metadata-filters '{\"top_folder\":{\"==\":\"BD\"}}'` returned only the BD document.\n\n## Common Pitfalls\n\n- HNSW is not the backend for edited docs. Use `flat` or `ivf`.\n- Compact indexes are read-only.\n- `build --docs` cannot preserve arbitrary app metadata; use `build-jsonl` for that.\n- Filesystem `event_time` and `source_type` are not present unless a source reader or JSONL producer emits them.\n- For Qwen\u002FE5-style asymmetric models, pass both document and query prompt templates.\n","# LEANN Indexing\n\nUse this when creating or maintaining indexes under `.leann\u002Findexes\u002F\u003Cname>\u002F`.\n\n## First Decision: Backend\n\nChoose by corpus size and whether content is edited in place.\n\n| Backend | Best use | Incremental `--docs` | Incremental `build-jsonl --incremental-by-id` |\n|---|---|---|---|\n| `flat` | Small\u002Fmedium edited corpora, exact recall | Add\u002Fremove\u002Fmodify in place | Add\u002Fremove\u002Fmodify in place |\n| `ivf` | Larger edited corpora, approximate FAISS search | Add\u002Fremove\u002Fmodify in place | Add\u002Fremove\u002Fmodify in place |\n| `hnsw` | Large mostly static corpus that fits in RAM | Any docs change triggers full rebuild | Add-only for non-compact indexes; changed\u002Fremoved rows rebuild |\n| `diskann` | Very large disk-backed corpus | Full rebuild | Full rebuild |\n| compact indexes | Archived\u002Fread-only indexes | No incremental | No incremental |\n\nDefault to `flat --no-recompute` for actively edited markdown\u002Fwiki\u002FCRM\u002Ftool-doc corpora. Default to `hnsw` for large mostly static corpora. Use `ivf` when flat is too slow but remove\u002Fmodify still matters.\n\n## Embeddings\n\nPick the embedding provider explicitly for reproducible builds. Examples:\n\n```bash\n--embedding-mode sentence-transformers \\\n--embedding-model facebook\u002Fcontriever\n```\n\nFor OpenAI-compatible embedding servers:\n\n```bash\n--embedding-mode openai \\\n--embedding-model \u003Cmodel-name> \\\n--embedding-api-base \u003Cbase-url> \\\n--embedding-api-key \u003Capi-key>\n```\n\nIf the server requires a non-empty placeholder key, pass one explicitly.\n\n## Build From Files\n\n```bash\nleann build docs --docs .\u002Fdocuments \\\n  --backend-name flat \\\n  --no-recompute \\\n  --embedding-mode sentence-transformers \\\n  --embedding-model facebook\u002Fcontriever\n```\n\n`build --docs` preserves file\u002Fpath metadata and stamps:\n\n- `indexed_at`\n- `created_at`\n- `modified_at`\n- `source_root`\n- `relative_path`\n- `folder_path`\n- `top_folder`\n\n### Folder Filters\n\nUse `--top-folder-depth` when the first path segment is only a wrapper folder.\n\n```bash\nleann build docs --docs \u002Fpath\u002Fto\u002Fdocs \\\n  --top-folder-depth 2 \\\n  --backend-name flat \\\n  --no-recompute \\\n  --embedding-mode sentence-transformers \\\n  --embedding-model facebook\u002Fcontriever\n```\n\nFor `Clients\u002FBD\u002Fproposal.md` with `--top-folder-depth 2`, metadata is:\n\n- `relative_path = Clients\u002FBD\u002Fproposal.md`\n- `folder_path = Clients\u002FBD`\n- `source_root = \u003Cdocs root>`\n- `top_folder = BD`\n\nThen search with:\n\n```bash\nleann search dropbox \"proposal\" \\\n  --metadata-filters '{\"top_folder\":{\"==\":\"BD\"}}'\n```\n\nDo not call this `category` in core LEANN. `top_folder` is factual filesystem metadata; app-specific category semantics belong in JSONL metadata or a source reader.\n\n## Build From JSONL\n\nUse `build-jsonl` when the caller already has one row per searchable object and needs exact metadata preserved.\n\n```jsonl\n{\"id\":\"workon\",\"text\":\"workon switch project repo context\",\"metadata\":{\"name\":\"workon\",\"group\":\"project\",\"tier\":\"hot\"}}\n```\n\n```bash\nleann build-jsonl code-mode-tools \\\n  --input .pi\u002Fcode-mode-leann\u002Ftools.jsonl \\\n  --text-field text \\\n  --metadata-field metadata \\\n  --id-field id \\\n  --incremental-by-id \\\n  --backend-name flat \\\n  --no-recompute \\\n  --embedding-mode sentence-transformers \\\n  --embedding-model facebook\u002Fcontriever\n```\n\nRules:\n\n- Every row needs a unique stable `--id-field`.\n- If `metadata.id` exists, it must match the top-level ID.\n- LEANN writes the stable ID into metadata as `id` and `source_document_id`.\n- `metadata.indexed_at` is ignored for rowhashes, so timestamp refreshes do not re-embed unchanged rows.\n\n## Rebuild And Watch\n\n```bash\nleann rebuild docs\nleann rebuild docs --force\nleann watch docs --interval 30\n```\n\n`rebuild` replays stored `build_config`, including docs paths, JSONL input, chunking fields, backend, embedding flags, `--top-folder-depth`, and `--incremental-by-id`.\n\nUse `--force` for a full rebuild. The content-addressed embedding cache keeps force rebuilds cheap when text is unchanged.\n\n## Code Indexes\n\n```bash\nleann build code --docs .\u002Fsrc --use-ast-chunking \\\n  --backend-name flat \\\n  --no-recompute \\\n  --embedding-mode sentence-transformers \\\n  --embedding-model facebook\u002Fcontriever\n```\n\nAST chunking supports Python, Java, C#, TS\u002FTSX\u002FJS. Other languages fall back to token chunking. Verify by reading `meta.json` after build.\n\n## Source Readers\n\nSpecialized readers emit richer source metadata, including `source_type`, `event_time`, `author`, and source-specific IDs when available.\n\n```bash\nleann index-browser chrome\nleann index-email\nleann index-calendar --max-count 1000\nleann index-imessage\nleann index-chatgpt --export-path ~\u002Fchatgpt-export.zip\nleann index-claude --export-path ~\u002Fclaude-export.zip\nleann index-wechat --export-dir ~\u002Fwechat-export\n```\n\nAll `index-*` commands share the same embedding flags as `leann build`.\n\n## Important Flags\n\nInput:\n\n- `--docs PATH [PATH ...]`\n- `--file-types .md,.pdf,.txt`\n- `--include-hidden` \u002F `--no-include-hidden`\n- `--top-folder-depth N`\n\nBackend:\n\n- `--backend-name {hnsw,diskann,ivf,flat}`\n- `--compact` \u002F `--no-compact`\n- `--recompute` \u002F `--no-recompute`\n- `--force`\n\nChunking:\n\n- `--doc-chunk-size N`, `--doc-chunk-overlap N`\n- `--code-chunk-size N`, `--code-chunk-overlap N`\n- `--use-ast-chunking`\n- `--ast-chunk-size N`, `--ast-chunk-overlap N`\n\nJSONL:\n\n- `--input FILE.jsonl`\n- `--text-field FIELD`\n- `--metadata-field FIELD`\n- `--id-field FIELD`\n- `--incremental-by-id`\n\nDefault `--docs` extensions include documents (`.txt`, `.md`, `.pdf`, `.docx`, `.pptx`), common code\u002Fconfig\u002Fmarkup files, and `.ipynb`. PDFs get PyMuPDF\u002Fpdfplumber first, then generic fallback.\n\n## Embedding Cache\n\nBuild-time embeddings use `~\u002F.leann\u002Fembed-cache.sqlite` by default.\n\n- Override: `LEANN_EMBED_CACHE_PATH=\u002Fpath\u002Fcache.sqlite`\n- Disable: `LEANN_EMBED_CACHE=0`\n- Force model revision key: `LEANN_EMBED_MODEL_REV=\u003Cstable-revision>`\n\nCache keys include canonicalization version, model, provider\u002Fmode, dimensions, normalization\u002Fpooling, prompt template, truncation settings, and model revision when known. Batch size and timeouts do not affect cache keys.\n\nBM25\u002FFTS5 is built and maintained by default. Do not pass a `--prebuild-bm25` flag in this fork.\n\n## Shipped: `--top-folder-depth`\n\nSHIP for the requested folder-filter change:\n\n- Added `leann build --top-folder-depth` in `packages\u002Fleann-core\u002Fsrc\u002Fleann\u002Fcli.py`.\n- Default is `1`, preserving current behavior.\n- With `--top-folder-depth 2`, `Clients\u002FBD\u002Fproposal.md` yields `top_folder = BD`.\n- Wired through build defaults, explicit option tracking, persisted `build_config`, rebuild, and watch replay.\n- Documented in `docs\u002Fconfiguration-guide.md` and `docs\u002FCHANGELOG.md`.\n- Verified with `uv run pytest tests\u002Ftest_cli_build_config.py tests\u002Ftest_indexed_at.py -q`: `22 passed`.\n- Verified with `uv run ruff check packages\u002Fleann-core\u002Fsrc\u002Fleann\u002Fcli.py tests\u002Ftest_cli_build_config.py tests\u002Ftest_indexed_at.py`.\n- Real CLI smoke confirmed `--metadata-filters '{\"top_folder\":{\"==\":\"BD\"}}'` returned only the BD document.\n\n## Common Pitfalls\n\n- HNSW is not the backend for edited docs. Use `flat` or `ivf`.\n- Compact indexes are read-only.\n- `build --docs` cannot preserve arbitrary app metadata; use `build-jsonl` for that.\n- Filesystem `event_time` and `source_type` are not present unless a source reader or JSONL producer emits them.\n- For Qwen\u002FE5-style asymmetric models, pass both document and query prompt templates.\n",{"name":5,"description":7},"zain-canonical","9a4916897c0f55a6f4593ac4fe22d4006894229d","skills\u002Fleann-index","https:\u002F\u002Fgithub.com\u002Fzm2231\u002FLEANN-fork","https:\u002F\u002Fgithub.com\u002Fzm2231","https:\u002F\u002Fmerchantry.systems","I set up AI around how businesses work, and build websites and small apps. Cofounder at Quoxient.","2026-07-21 07:54:37",{"source":8,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add zm2231\u002FLEANN-fork --skill leann-index",{"tool":38,"command":39},"bunx","bunx skills add zm2231\u002FLEANN-fork --skill leann-index",{"tool":41,"command":42},"pnpm","pnpm dlx skills add zm2231\u002FLEANN-fork --skill leann-index","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fzm2231\u002FLEANN-fork\u002Fblob\u002Fzain-canonical\u002Fskills\u002Fleann-index\u002FSKILL.md",true,[],[47,54,60,66,72,78],{"skillNo":48,"name":49,"nameZh":6,"description":50,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":51,"categories":52,"keywords":53,"publishedAt":19},"01M756X633PN3NTPGG24VVJHD1","leann-search","Search LEANN indexes through CLI, MCP, or Python. Use for `leann search`, `ask`, `react`, metadata filters, hybrid BM25\u002Fvector search, temporal queries, sparse-filter diagnostics, result diversification, context windows, facets, and batched multi-search.","https:\u002F\u002Fgithub.com\u002Fzm2231\u002FLEANN-fork\u002Fblob\u002Fzain-canonical\u002Fskills\u002Fleann-search\u002FSKILL.md",[],[],{"skillNo":55,"name":56,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":57,"categories":58,"keywords":59,"publishedAt":19},"01M756X2CFEH8GZFWX0DQSMJSB","whatsapp","https:\u002F\u002Fgithub.com\u002Fzm2231\u002FLEANN-fork\u002Fblob\u002Fzain-canonical\u002Fpackages\u002Fleann-sources\u002Fsources\u002Fmessaging\u002Fwhatsapp\u002FSKILL.md",[],[],{"skillNo":61,"name":62,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":19},"01M756X13J8DAWR9PJARESXS6W","wechat","https:\u002F\u002Fgithub.com\u002Fzm2231\u002FLEANN-fork\u002Fblob\u002Fzain-canonical\u002Fpackages\u002Fleann-sources\u002Fsources\u002Fmessaging\u002Fwechat\u002FSKILL.md",[],[],{"skillNo":67,"name":68,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":69,"categories":70,"keywords":71,"publishedAt":19},"01M756WYN04ES3Y3A5J872CD9E","imessage","https:\u002F\u002Fgithub.com\u002Fzm2231\u002FLEANN-fork\u002Fblob\u002Fzain-canonical\u002Fpackages\u002Fleann-sources\u002Fsources\u002Fmessaging\u002Fimessage\u002FSKILL.md",[],[],{"skillNo":73,"name":74,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":19},"01M756WWR6JQ82R18JR7RYWTNN","claude-export","https:\u002F\u002Fgithub.com\u002Fzm2231\u002FLEANN-fork\u002Fblob\u002Fzain-canonical\u002Fpackages\u002Fleann-sources\u002Fsources\u002Fimports\u002Fclaude-export\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M756WVP3AZMVWJQF4Z7P4Y3W","chatgpt-export","https:\u002F\u002Fgithub.com\u002Fzm2231\u002FLEANN-fork\u002Fblob\u002Fzain-canonical\u002Fpackages\u002Fleann-sources\u002Fsources\u002Fimports\u002Fchatgpt-export\u002FSKILL.md",[],[]]