wiki-ingest files supplied source material into an Obsidian vault without changing the source. It agrees a budget with you, hashes each source, classifies it (code, research, decision, conversation, web reference, dataset or media), and writes pages only where the source adds durable knowledge.
Source and claim ledgers record authority, freshness and independence. Everything lands in one hashed transaction you review before it is applied.
What You Get
- Source summaries and linked canonical pages with their sources recorded
- Create-only raw captures and updated source and claim ledgers
- Index or map-of-content entries, a log entry and a refreshed hot cache
- A preview with inputs, budget used, claim assessments and skipped items
- An operation ID and exact changed paths after apply
Common Use Cases
- Filing a paper, article or transcript with citations into a vault
- Ingesting a bounded batch of files from an inbox folder
- Drafting a cited glossary or consolidating overlapping notes with reviewed scope
What Makes It Unique
It treats every source as untrusted data and ignores embedded instructions, will not fetch a URL without domain consent, and refuses to claim extraction it did not do for PDFs, images, audio or OCR. Each canonical page create is tied to an index update, and a changed remote source gets a new immutable capture in place of an overwrite.
Under The Hood
The skill is instructions that point to the plugin’s provenance and transaction references. It reads files in the vault and may fetch web pages, but only after you approve the destination domains and a request budget, and it stops if redirects leave that scope. It runs the plugin’s Python core script through the shell to inspect and apply a workflow bundle after you approve the plan hash, and that step writes pages, ledgers, raw captures, the log and the hot cache. A Git checkpoint is a separate optional command. Helper agents return drafts only.