Knowledge Vault

Research Intake Protocol

3 min read

Research Intake Protocol

Standardized process for integrating external research materials into the knowledge vault. Established through Phases 1-6 of the research-lab-materials integration (January-February 2026).

1. Capture with Frontmatter

Every incoming file must have YAML frontmatter from the start:

title: Descriptive Title
created: YYYY-MM-DD
type: research | reference | pattern
source: origin path or URL
content-type: academic-paper | ebook-extract | technical-guide | org-blueprint
tags:
  - category
  - technology
  - imported-YYYY-MM

2. Classify by Priority

PriorityCriteriaAction
P1-CriticalAP-generated blueprints, core ML/AI, security architectureIntegrate immediately
P2-ImportantLeadership references, active research dependencies, infrastructureIntegrate in next batch
P3-ReferenceAcademic papers, programming fundamentals, supplementary materialIntegrate when bandwidth allows
Stub/EmptyFiles under 500 bytes with no extractable contentSkip with documentation

3. Check for Duplicates

Before integrating, verify the content doesn’t already exist in the vault:

  • Search by title keywords in existing vault files
  • Compare file sizes for identical source files (content hashing — see Data Deduplication Architecture)
  • Check the INTEGRATION-LOG.md for prior processing

4. Read Files Before Trusting Filenames

At least 9 of 137 misc files had content that didn’t match their filename. Always open and read the first 20-30 lines to verify actual content before classifying or naming the vault entry.

Related papers or documents should be merged into single vault files to avoid thin duplicate pages:

  • Academic papers from the same research program (e.g., 4 PRA papers from Lao/Cohen)
  • Multiple sources on the same narrow topic (e.g., 3 ergodicity sources)
  • Book stubs + reviews of the same work

This creates stronger, more interconnected entries. See Cross-Cutting Theme Clusters for the organizational principle.

Every integrated file must link to at least two other vault pages. This is the Minimum 2 Wikilinks requirement. Strategies:

  • Link to the relevant index page (research/index, references/methodology/index, etc.)
  • Link to thematically related vault content
  • Link to patterns or projects that build on this research

7. Validate After Each Batch

Run npm run validate:links after every integration batch to catch:

  • Broken links from typos in wikilink paths
  • New orphan pages that need incoming links
  • Total file count changes

8. Update INTEGRATION-LOG.md

After each phase:

  • Add entries with source file, destination, status, and notes
  • Bump the log version number
  • Update the processing summary table
  • Document any new decisions in the Decision Log section

Lessons Learned

These are hard-won insights from processing 198 files across 6 phases:

LessonImpact
Filename-content mismatches are common (~7% of files)Must read before trusting names
PDF-encoded text files appear as garbled UnicodeRequire special extraction tooling
Stub files waste integration timeFilter by size early (<500 bytes = skip)
Merging related papers creates stronger entriesReduces file count, increases depth
Index pages are the highest-leverage orphan fixOne edit can de-orphan 10-30 pages
French/non-English sources need language taggingMay have limited utility depending on audience
Background agents can’t write filesMust write from main context when using AI tooling

Process Checklist

For each file being integrated:

  • Read actual content (don’t trust filename)
  • Check for duplicates in vault and INTEGRATION-LOG
  • Determine priority (P1/P2/P3) and content type
  • Identify merge candidates among related files
  • Create vault file with complete frontmatter
  • Add minimum 2 wikilinks
  • Ensure at least one incoming link exists (from index page)
  • Run validation
  • Update INTEGRATION-LOG.md

Graph View

Backlinks