Knowledge Bank Ingest

Ingest external knowledge sources into the knowledge bank. Two modes serve different depths of engagement:

• Quick mode: Summarize → agree on doc type → file to KB. For sources you want to preserve but don't need to deeply study.
• Study mode: Theme-by-theme deep dive with online research → personal notes → multi-layer KB output → optional synthesis. For sources you want to actively learn from and potentially share.

Knowledge Bank Location: Read from

~/.claude/plugins/config/second-brain/config.json

Philosophy: The discussion between you and the user shapes what gets preserved — this is not a summarizer, it's a collaborative knowledge integration tool. A single source can touch multiple KB pages.

Invocation

/second-brain:kb-ingest /path/to/article.md
/second-brain:kb-ingest https://example.com/interesting-post

Step 1: READ

Read the source document (file path → Read tool, URL → WebFetch, inline → accept directly).

For large sources (>100KB), apply the distillation guide.

Identify the source's major themes, key concepts, and actionable patterns.

Step 2: MODE SELECT

After reading, use the AskUserQuestion tool to let the user choose their engagement mode:

• Quick — Summarize key points, agree on doc type, file to KB
• Study — Walk through theme by theme with online research and personal notes, then create multi-layer KB output

Step 3: ENGAGE

Quick Mode

1. Summarize the source's main contributions (3-5 bullet points)
2. Propose what to preserve and as what type (concept / component / best-practice / multiple)
3. Use the AskUserQuestion tool to confirm: What resonates? What project context? Anything to add from their own experience?
4. Proceed to Step 4 (CREATE) once agreed

Study Mode

Walk through the source theme by theme. For each theme:

1. Present — explain the core idea, key examples, and why it matters
2. Explore — research online (WebSearch/WebFetch) to find 2-3 relevant pieces that enrich the theme: related work, counter-arguments, real-world applications. The goal is to add context, not exhaustively survey. See study-mode-guide.md for depth guidance and a walkthrough example.
3. Pause — let the user note their takeaways in their own words
4. Capture — record the user's personal insights, reframings, and connections to existing knowledge. The user often synthesizes more sharply than the source — preserve their framing, not just the article's.
5. Repeat for the next theme until the source is covered

The user's notes are first-class output — they're not just "discussion", they're the learning record that becomes the study docs in Step 4.

Step 4: CREATE

Quick Mode Output

Generate concept/component/best-practice doc(s) as in the ingest guide. Target 5-8 WikiLinks.

Study Mode Output — Three Layers

Study mode produces three layers of documentation, each serving a different future use:

projects/{project}/sources/

projects/{project}/study/

projects/{project}/study/

Plus the standard concept/component/best-practice docs distilled from the study.

Frontmatter for all docs:

---
title: Document Title
aliases: [Alt 1, Alt 2]
tags: [category, topic]
type: concept|component|best-practice|study|source
created: YYYY-MM-DD
modified: YYYY-MM-DD
project: {project}
source-type: article|document|url|gist
ingested-from: /path/to/source.md or https://url
source-doc: "[[Source Document Title]]"
---

Cross-references: Run

search_cross_references.sh "keyword" [project]

Obsidian Syntax: When obsidian skills are available, invoke

/obsidian:obsidian-markdown

Updating existing docs: If new knowledge connects to or supersedes existing KB pages, update them — add cross-references, revise outdated claims, note where new data strengthens or challenges existing content.

Step 5: SYNTHESIZE (optional, after Study mode)

After Study mode CREATE is complete, use the AskUserQuestion tool to ask what synthesis the user wants:

• Blog post — Synthesize into a listicle-style blog post following the blog synthesis guide
• Slack message — Concise narrative prose (~150 words), no bullets, links at bottom
• Skip — No synthesis needed, KB docs are sufficient

The synthesis draws from the digest + learning notes + user's personal reframings — not the raw source. The user's voice and analytical angle are the value-add over the original article.

Key principle: The user owns their voice. Present drafts, but expect the user to rewrite key passages (especially openings). Polish what they give you — don't ghostwrite.

Save synthesis outputs to the session docs folder:

• docs/polished-blog-YYYYMMDD/

  — original draft, polished version, changelog, feedback

• docs/polished-slack-YYYYMMDD/

  — same structure for Slack

Step 6: INTEGRATE

After creating/updating documents:

1. Regenerate index (so new docs appear in the unified catalog):

source skills/common/generate_index.sh
generate_index "$KB_PATH"

2. Append operation log (audit trail for KB changes):

source skills/common/obsidian_helpers.sh
append_kb_log "$KB_PATH" "ingest" "kb-ingest" "Ingested [source]. Created: [docs]. Updated: [docs]"

3. Verify created documents:

./skills/session-recap/scripts/count_wikilinks.sh document.md 5
./skills/session-recap/scripts/validate_obsidian_syntax.sh document.md

Batch Ingest

For multiple sources: "Ingest all .md files in /path/to/folder/"

• READ all sources first
• ENGAGE as a batch — present combined takeaways, let user prioritize
• CREATE docs per source (or merged if topics overlap)
• INTEGRATE once at the end

Resources

• KB Schema — Unified conventions for all KB documents
• Ingest Guide — Source types, classification, quality standards
• Study Mode Guide — Theme walkthrough example, explore depth guidance, capturing user notes
• Blog Synthesis Guide — How to create blog posts from studied material
• Distillation Guide — For large sources

Completion Criteria