Meditation Facilitation Skill

Interactive knowledge consolidation. Dream is the MRI; meditation is the surgery.

Meditation vs Dream

Dimension	Dream (Automated)	Meditation (Interactive)
Trigger	Command / schedule	User says "meditate"
Depth	Structural validation	Semantic understanding
Output	Health metrics JSON	Enriched memory files
Analogy	X-ray machine	Surgeon with the X-ray
Decision-making	None (report only)	Yes (create, update, prune)

Session Types and When to Use Each

Type	Duration	Trigger Signal	Output
Quick Reflect	5-10 min	End of coding session	Session record in episodic/
Topic Deep-Dive	15-30 min	"I learned something important"	Updated SKILL.md or new insight
Full Meditation	30-60 min	End of sprint/feature	Memory file updates + synthesis
Self-Actualization	45-90 min	Monthly or architecture change	Comprehensive assessment report

The 4 R's — Expanded

Review — What happened this session?

• List concrete actions taken (files created, bugs fixed, features added)
• Identify surprising discoveries or unexpected connections
• Note what was harder or easier than expected and why
• Quality check: If the review is all "I did X, Y, Z" with no reflection, it's too shallow

Relate — How does this connect to existing knowledge?

• Which existing skills/instructions does this session's work touch?
• Are there cross-project patterns? (→ global knowledge candidate)
• Did this session reveal a gap in the architecture?
• Quality check: At least one connection identified, or explicitly note "isolated learning"

Reinforce — What's worth remembering long-term?

Filter aggressively. Not everything is worth persisting:

Signal	Worth Remembering?	Store Where
"I'll forget this by next week"	Yes	SKILL.md or .instructions.md
"This was non-obvious"	Yes	Global insight (GI-*)
"I keep rediscovering this"	Definitely	.instructions.md (procedural)
"This was specific to today"	No	Session record only
"Everyone knows this"	No	Don't store

Record — Persist the right things in the right places

Learning Type	Location	Format
Repeatable process	.instructions.md	Step-by-step procedure
Interactive workflow	.prompt.md	Guided conversation template
Domain expertise	skills/*/SKILL.md	Reference knowledge
Cross-project pattern	global-knowledge/patterns/	GK-* pattern
Timestamped observation	global-knowledge/insights/	GI-* insight
Session narrative	.github/episodic/	Dated session record

Facilitation Techniques

Socratic Probing

Don't just list what happened — ask why it matters:

• "What would you do differently next time?"
• "Is this a one-time fix or a recurring pattern?"
• "If you saw this in another project, would you recognize it?"

Diminishing Returns Detection

Signal	Action
Repeating the same insight in different words	Move on
No new connections after 3 attempts	Close this topic
User energy dropping	Wrap up with quick record
Tangent detected	Bookmark for separate session

Depth Calibration

• Too shallow: "I learned about testing" → probe for specifics
• Right depth: "Property-based testing caught an edge case our unit tests missed because..."
• Too deep: 30 minutes on one function's implementation → zoom out to pattern level

Session Record Template

Store in .github/episodic/YYYY-MM-DD-topic.md:

# Meditation: [Topic]
**Date**: YYYY-MM-DD
**Type**: Quick Reflect | Deep-Dive | Full | Self-Actualization
**Duration**: X minutes

## Focus
What this session was about.

## Key Learnings
- Learning 1 (→ stored in [location])
- Learning 2 (→ connected to [skill/instruction])

## Updates Made
- File X: added/updated Y
- Synapse: connected A → B

## Open Questions
- What still needs investigation?

Synapses

See synapses.json for connections.