Manage Thinking Topics

Manage Thinking Topics

Human-in-the-loop manager for the incubation loop topic registry. Handles the full lifecycle: seeding new questions into analysis, reviewing current status, graduating converged conclusions to permanent notes, and retiring topics no longer worth pursuing.

Companion skill: .claude/skills/incubation-loop/SKILL.md — the analytical engine that processes active topics.

State Dependencies

Modes

Invoke with one of four modes. If no mode is provided, default to review.

Mode: review

Purpose: Show the full topic landscape at a glance - what's active, what needs crystallization, what's done.

Step 1: Read Registry

date '+%Y-%m-%d'

Read Brain/05-Meta/Thinking/THINKING-REGISTRY.md.

Step 2: For Each Active Topic - Load Summary

For each topic with status: active, read its thinking file. Extract:

• run_count - how many iterations completed
• last_run - date of last iteration
• Current leading hypothesis + confidence score
• Next move in rotation (run_count mod 6)
• Any open questions flagged

Step 3: For Each Converged Topic - Load Summary

For each topic with status: converged, read its thinking file. Extract:

• The converged conclusion (Current Best Answer from last run)
• Confidence score at convergence
• Run count at convergence

Step 4: Present Status Report

Output a structured report:

## Thinking Topics: Status Review [DATE]

### Active Topics (N)
| Topic | Runs | Last Run | Leading Hypothesis | Confidence | Next Move |
|-------|------|----------|-------------------|------------|-----------|

### Converged — Ready for Crystallization (N)
| Topic | Runs | Conclusion Summary | Confidence |
|-------|------|-------------------|------------|

### Crystallized — Done (N)
| Topic | Permanent Note | Crystallized |
|-------|---------------|-------------|

### Summary
- [N] topics in active analysis
- [N] topics ready for crystallization (action needed)
- [N] topics fully crystallized
- Next incubation-loop run: processes all active topics

If there are converged topics, highlight: "[N] topics ready for crystallization — run /manage-thinking-topics crystallize [slug] to graduate each one."

Mode: seed

Purpose: Add a new question to the incubation loop.

Argument: topic slug (e.g., agent-memory-architecture)

Step 1: Gather Requirements

If the topic slug is not provided, ask for it. Then gather:

1. Central question - The exact question being analyzed. Precise framing matters - vague questions stay vague.
2. Why it matters - 1-2 sentences on stakes or relevance.
3. Initial hypotheses - At least 2 competing hypotheses with rough initial confidence percentages (must sum to ~100%).
4. Known evidence - Any notes or sources already known to be relevant (optional).
5. Constraints/assumptions - What's taken as given? What's out of scope? (optional)

Step 2: Validate Slug

Check that Brain/05-Meta/Thinking/[topic-slug].md does NOT already exist. If it does, warn and stop.

Check that the slug is not already in the registry. If it is, warn and stop.

Step 3: Create Topic File

date '+%Y-%m-%d'

Create Brain/05-Meta/Thinking/[topic-slug].md:

---
created: YYYY-MM-DD
updated: YYYY-MM-DD
created_by: claude-sonnet-4-6
updated_by: claude-sonnet-4-6
agent_version: 01.25
topic: "[exact question being analyzed]"
run_count: 0
last_run: null
status: active
---

# Thinking: [Topic Title]

## Central Question
[The exact question. Precise framing matters - ambiguous questions stay ambiguous.]

## Why This Question Matters
[1-2 sentences on stakes or relevance]

## Initial Hypotheses
| Hypothesis | Initial Confidence | Basis |
|-----------|-------------------|-------|
| H1: [statement] | X% | [prior knowledge or intuition] |
| H2: [statement] | X% | [prior knowledge or intuition] |

## Known Evidence (Pre-Run)
[Any notes or sources already known to be relevant, or "None identified yet."]

## Constraints and Assumptions
[What are you taking as given? What's out of scope?]

---
*Analytical runs will be appended below by the incubation loop*

Step 4: Add to Registry

Read Brain/05-Meta/Thinking/THINKING-REGISTRY.md.

Add a row to the Active Topics table:

| [topic-slug] | [central question] | active | 0 | null |

Update updated and updated_by in frontmatter.

Step 5: Confirm

Report:

• Topic file created at Brain/05-Meta/Thinking/[topic-slug].md
• Registry updated - topic is now active
• First incubation-loop run will apply: ACH Audit (move 0)
• Estimated convergence: 4-6 runs (~4-6 days at daily schedule)

Mode: crystallize

Purpose: Graduate a converged topic — either as an atomic permanent note (Brain/02-Permanent/, graph-integrated) or as a terminal synthesis report (Brain/05-Meta/Reports/, excluded from semantic indexing). Choose based on whether the conclusion is generative (future reasoning will reuse it) or terminal (snapshot of what was concluded).

Argument: topic slug of a converged topic

If no slug is provided: read the registry, present the converged topics (with one-line conclusions), and ask which to crystallize before proceeding. If prior session context points at a specific topic or pair (e.g., a just-completed analysis that named them), recommend it as the first option. Accept common misspellings of the mode name (e.g., "crystalize").

Step 1: Load Converged Topic

Read Brain/05-Meta/Thinking/[topic-slug].md.

Verify status: converged. If not converged, warn and stop.

Extraction strategy (prefer the canonical block, fall back gracefully):

1. First, look for ## Final Synthesis - a canonical block appended when the loop transitions a topic to converged. If present, this is the authoritative source. Extract from it:

   • Conclusion (the substantive answer)
   • Confidence and leading hypothesis
   • Key evidence (already curated)
   • Remaining uncertainty
   • Related topics (if any)

2. If no Final Synthesis block exists (legacy topics from before the standardized format), read only the last 200 lines of the file rather than the full thinking journal (some are 5,000+ lines). Extract:

   • The central question (from frontmatter)
   • The most recent "Current Best Answer" section
   • The most recent "Updated Hypotheses" table for confidence scores
   • Any "Open Questions" flagged in the final run

3. Read the full file only if the last 200 lines do not contain a "Current Best Answer" or equivalent conclusion.

Step 2: Search for Existing Notes

python3 resources/local-brain-search/search.py "[central question keywords]" --limit 5 --mode spreading 2>/dev/null

Check if a permanent note already exists covering this conclusion. If so, note the overlap and ask whether to update it or create a new note.

Step 3: Choose Destination

Before drafting, decide the destination based on the nature of the conclusion:

[APPROVAL GATE] Propose the destination and ask the user to confirm or override:

"Recommend: [destination] because [reason]. Confirm, or override to [other destination]?"

Synthesis reports do not participate in connection-finder or semantic search (the Reports/ folder is excluded from LBS indexing). They are visible in Obsidian and can link OUT to permanent notes via [[wiki-links]], but permanent notes should rarely link IN to reports.