Knowledge Capture

Input: $ARGUMENTS

Note: /mem exists for mental models. This skill is for personal knowledge management — capturing insights, decisions, patterns, and facts in a way your future self can retrieve and use.

Step 1: Identify What’s Worth Capturing

Not everything deserves a knowledge entry. Filter the input through these criteria:

1. Insight: Something you understood for the first time or in a new way
2. Decision + rationale: A choice you made and WHY (the “why” decays fastest)
3. Pattern: A recurring observation across multiple instances
4. Hard-won fact: Something that took effort to find or verify
5. Procedure: A sequence of steps you figured out and will need again
6. Mistake + lesson: An error and what it taught you

CAPTURE TYPE: [insight / decision / pattern / fact / procedure / mistake]
WORTH CAPTURING BECAUSE: [why this will matter later]
EXPIRY RISK: [will this become outdated? If so, when?]

SKIP: If the input doesn’t pass any of these filters, tell the user — not everything needs to be captured. Over-capturing creates noise that drowns signal.

Step 2: Choose Storage Format

Match the format to the knowledge type:

Write the entry in the chosen format now. Rules:

• Write for your future self who has forgotten the context
• Lead with the conclusion, not the backstory
• Include ONE concrete example
• Keep it under 200 words

--- KNOWLEDGE ENTRY ---
[formatted entry here]
--- END ---

Step 3: Create Retrieval Structure

Knowledge you can’t find is knowledge you don’t have. Design retrieval paths:

1. Tags: 2-5 keywords your future self would search for
   • Include both the topic and the domain
   • Include at least one “when would I need this” tag
2. Hierarchy: Where does this fit in your existing knowledge structure?
   • Parent category
   • Related entries (if any)
3. Links: What other knowledge does this connect to?
   • What does this depend on?
   • What depends on this?

TAGS: [tag1, tag2, tag3, ...]
CATEGORY: [parent] > [subcategory]
LINKS TO: [related entries or topics]
RETRIEVAL TRIGGER: [situation where you'd need this]

Step 4: Write in Future-Self-Friendly Format

Review the entry from Step 2 and check:

1. Context independence: Would this make sense with zero memory of today?
2. Actionability: Can future-you act on this without further research?
3. Scannability: Can the key point be grasped in under 10 seconds?
4. Specificity: Are there concrete details, not just vague principles?

Rewrite if any check fails. The most common failure mode is writing for present-you who has full context instead of future-you who has none.

Step 5: Schedule Review

Knowledge entries decay. Set a review schedule:

NEXT REVIEW: [date or trigger condition]
REVIEW ACTION: [verify / update / archive / delete]

Step 6: Prune Outdated Entries

If the user is reviewing existing knowledge (not capturing new):

1. For each entry, ask: “If I discovered this today, would I capture it?”
2. If NO: archive or delete
3. If YES but it’s stale: update with current understanding
4. If YES and current: keep, confirm review date

Signs an entry is stale:

• The technology/tool it references has changed
• Your understanding has evolved past it
• You’ve never retrieved it in 12+ months
• It contradicts something you now know to be true

PRUNING RESULT:
- Kept: [count]
- Updated: [count]
- Archived: [count]
- Deleted: [count]

Integration

Use with:

• /lr -> Research a topic, then capture key findings with /memk
• /rca -> After root cause analysis, capture the lesson
• /curd -> Capture knowledge milestones during a learning path
• /dcp -> After a decision, capture the rationale before you forget it