Ingest URLs, files, and documents into gbrain with entity extraction and cross-linking.
npx skills add https://github.com/tapway/shogun-os --skill gbrain-ingest-tapway
Please help me install this Agent Skill. Skill: gbrain-ingest Source: https://github.com/tapway/shogun-os/tree/main/skills/gbrain/gbrain-ingest Command: npx skills add https://github.com/tapway/shogun-os --skill gbrain-ingest-tapway
references/, projects/, or decisions/. - Entity extraction and cross-linking: Scans content for people, companies, and projects, resolves them against existing brain pages, and creates back-links from every mentioned entity. - Bulk import support: Imports 500+ files via gbrain import with a subagent batching pattern that avoids out-of-memory crashes on large directories. - Use Case: You receive a PDF contract from a vendor. The Skill extracts the text, identifies entities like the company and signatory, writes the page under references/contracts/, and links it back to the existing company and people pages. ## Quick Start Import this PDF contract into the brain, extract the key entities, and cross-link it to the relevant company and people pages.High-intent search queries and answers about installing and using this skill.
Use `gbrain import <dir> --no-embed --workers N` for bulk imports. For 500+ files, split files into batch directories of about 200 files using hardlinks, then run subagents with `--workers 1` to avoid exhausting system RAM.
Extract the PDF text with OCR tools like pymupdf or marker-pdf, identify key entities such as people and companies, then write the page under the correct brain path and add back-links from each entity page to the new page.
The default `--workers 8` setting can consume 14GB+ RSS on thousands of files due to connection pools and JIT heap, not the file data itself. Fix it by batching files into smaller directories and importing with `--workers 1` per subagent.
The gbrain import walker explicitly skips symlinks as a defense against symlink cycles, and there is no override flag. Use hardlinks via `os.link` or plain file copies when building batch directories.
Avoid it for media content needing transcripts or OCR (use a media ingest skill), external articles needing summarization (use an idea ingest skill), and original user thoughts, which belong in a capture workflow instead.
RonanCodesExtract web article content into markdown with YAML frontmatter.
garrytanEnforces brain-first research workflow before falling back to web search.
SigmacodeatDetermine filing locations for new brain pages using an active taxonomy schema.
understudy-aiSummarize text from URLs, local files, and video transcripts.