Add wiki-warehouse skill: cold storage for raw documents#32
Merged
Conversation
The vault-side operating layer for the two-repo knowledge system. Raw documents (PDFs, ebooks, scans) go into a separate private knowledge-warehouse repo via its intake.py (hash -> extract text, OCR fallback for scans -> manifest); the vault gets only a lean Source note with a summary, key excerpts, and a content-hash (doc_id) pointer — keeping originals and full text out of the vault so it stays fast and git-diffable. - skills/wiki-warehouse/SKILL.md — /ingest (drive intake.py, push warehouse, write the pointer note) and /warehouse-audit (reconcile vault doc_id pointers against the warehouse manifest: dangling / drifted / orphaned). - references/warehouse-schema.md — the vault<->warehouse join contract: doc_id semantics, Source-note frontmatter fields, pointer-integrity buckets. - Registered in marketplace.json + README (skills table, structure tree, wiki-system section). The warehouse repo itself (Hefrock/knowledge-warehouse, private) holds intake.py and was verified end-to-end on born-digital PDF, plaintext, dedupe, and the scanned/OCR-unavailable path. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01D3gQfyqFnxTqnm455xwPyX
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds the vault-side operating layer for a two-repo knowledge system: a lean Obsidian index + a separate private warehouse for raw documents.
skills/wiki-warehouse/SKILL.md—/ingest(drive the warehouse'sintake.py, push the original, then write a lean vault Source note) and/warehouse-audit(reconcile vaultdoc_idpointers against the warehouse manifest — dangling / drifted / orphaned).references/warehouse-schema.md— the vault↔warehouse join contract:doc_id(sha256 content hash) semantics, the Source-note frontmatter fields, and the pointer-integrity buckets.marketplace.jsonand the README (skills table, structure tree, wiki-system section).How it works
Raw documents (PDFs/ebooks/scans) go into a separate private
Hefrock/knowledge-warehouserepo via itsintake.py: hash the content → extract text (born-digital by default, OCR fallback for scans) → update a content-hash manifest. The vault gets only a summary + key excerpts + adoc_idpointer — never the binary, never the full text — so it stays fast and its git history stays small. The join key is a content hash, not a path, so the warehouse can be reorganized without breaking vault pointers (and/warehouse-auditrepairs any drift).Design decisions (confirmed with the user)
Mostly born-digital documents (OCR is the exception path, auto-detected); summary + excerpts only in the vault (full text stays warehoused); occasional one-at-a-time intake (a manual command, not scheduled automation); warehouse is private (may hold sensitive/PHI-adjacent material).
Test plan
intake.pywas built and verified end-to-end in a sandbox: born-digital PDF extraction (PyMuPDF), plaintext, content-hash dedupe (re-ingest is a no-op), and the scanned/OCR-unavailable path (detects near-zero text, records the doc with a clear "install ocrmypdf and re-run" warning rather than failing).marketplace.jsonre-validated as JSON after adding the plugin entry.SKILL.mdfrontmatter validated (name + description present).Note: the warehouse repo (
Hefrock/knowledge-warehouse, private) is already initialized withintake.py, the manifest, theraw/+text/structure, and its own README — this PR is only the vault-side skill in agent-skills.🤖 Generated with Claude Code
https://claude.ai/code/session_01D3gQfyqFnxTqnm455xwPyX
Generated by Claude Code