incubation-loop

name	incubation-loop
description	Autonomous iterative thinking loop - processes active topics using rotating analytical moves (ACH, Bayesian updating, steelmanning, cross-domain bridging, implication checks) and persists reasoning state across scheduled runs
automation	autonomous
schedule	0 7 * * *
allowed-tools	Read, Write, Grep, Glob, Bash, WebSearch, Skill
user-invocable	true
argument-hint	[topic-slug] (optional - processes all active topics if omitted)
metadata	{"version":"1.1","updated":"2026-07-03T00:00:00.000Z","changelog":["1.1: Registry size discipline (Step 8) — single one-line last_incubation_run overwritten in place, no _prior_* history keys, no analytical narrative in the registry; converged rows capped to a one-sentence conclusion + pointer to the topic file's Final Synthesis","1.0: Initial version"]}

Incubation Loop

ℹ️ First, set expectations: before anything else, print one short line with this skill's version and its most recent change — the top entry of metadata.changelog above — e.g. incubation-loop vX.Y — recent: <summary>. Then proceed.

Autonomous thinking engine. Each scheduled run advances all active thinking topics by one analytical move, persisting reasoning state across runs until convergence.

Purpose

Continuous intellectual iteration on open questions. Uses a rotating set of research-validated analytical moves (ACH, Bayesian updating, dialectical steelmanning, cross-domain bridging) to build toward well-grounded conclusions - without requiring human presence in each cycle.

Design principle: Each run does one move per topic. Depth accumulates across runs. No single run tries to "solve" the question.

State Dependencies

Source	Location	Read	Write	Description
Thinking Registry	Brain/05-Meta/Thinking/THINKING-REGISTRY.md	✓	✓	Active topics + status
Thinking Files	Brain/05-Meta/Thinking/[topic-slug].md	✓	✓	Per-topic reasoning journal
Local Brain Search	resources/local-brain-search/	✓		Semantic search for KB evidence
Permanent Notes	Brain/02-Permanent/	✓		Primary evidence source
Session Changelogs	Brain/05-Meta/Changelogs/		✓	Run log

Prerequisites

• Registry file exists at Brain/05-Meta/Thinking/THINKING-REGISTRY.md
• At least one active topic seeded (see Seeding section below) - or none, in which case the Starvation Floor (Step 1a) self-seeds one from the watched domains
• Local Brain Search index up-to-date

Composes

• /domain-watch - invoked by the Starvation Floor (Step 1a) when the active queue is empty, to formulate and activate a new topic from the domains under surveillance. Called by its unversioned name so its fixes propagate.

---

Process

Step 1: Get Date and Load Registry

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

Read Brain/05-Meta/Thinking/THINKING-REGISTRY.md. Parse all topics with status: active.

If registry does not exist, create it:

---
created: YYYY-MM-DD
updated: YYYY-MM-DD
created_by: claude-sonnet-4-6
updated_by: claude-sonnet-4-6
agent_version: 01.25
---

# Thinking Registry

Active questions under continuous analysis.

| Topic Slug | Central Question | Status | Runs | Last Run |
|------------|-----------------|--------|------|----------|

If no active topics found, do NOT exit - run the Starvation Floor (Step 1a) so the thinking engine never idles.

Step 1a: Starvation Floor (only if zero active topics)

The thinking engine must never sit idle. If Step 1 found no active topics:

1. Invoke /domain-watch (cross-skill). Its Starvation Floor guarantees at least one new topic is activated into the Thinking Registry, formulated from the domains under surveillance.
2. Re-read Brain/05-Meta/Thinking/THINKING-REGISTRY.md and parse status: active topics again.
3. If one or more active topics now exist, continue to Step 2 and process them this run.
4. Only if /domain-watch still produced no topic (e.g. no domains configured), log to changelog and exit cleanly.

This makes topic generation autonomous: when human-seeded questions run out, the agent proactively formulates its own next question from the watched domains rather than going quiet. Crystallization of any resulting conclusion stays manual (human judgment).

Step 2: For Each Active Topic - Load State

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

Parse frontmatter fields:

• run_count (required) - how many iterations completed; drives move rotation
• last_run (required) - date of last iteration
• status (required) - active / converged / crystallized / retired
• last_move (optional, written by loop) - which move was applied in the most recent run; informational
• convergence_count (optional, written by loop) - number of consecutive runs at or near convergence criteria; used by framework-exhaustion detection in Step 7
• topic (set at seeding) - the central question being analyzed

Parse from body:

• current_hypotheses - list with confidence scores (from the most recent run's "Updated Hypotheses" table)
• Last run's "Current Best Answer" and "Open Questions"

Step 3: Determine Next Thinking Move

Rotate through this sequence. Position = run_count mod 6:

Position	Move	Research Basis
0	ACH Audit	Heuer (CIA) - Analysis of Competing Hypotheses
1	Bayesian Update	Information theory - explicit confidence tracking
2	Steelman Opposition	Socratic dialectic - steel-man the opposite
3	Cross-Domain Bridge	Consilience method - what does an unrelated field say?
4	Implication Check	Falsificationism (Popper) - if true, what follows? Is that true?
5	Assumption Audit	Intelligence SAT - which premises are load-bearing and shakiest?

Step 4: Run KB Search for Evidence

Use Local Brain Search to find relevant notes:

cd $PROJECT_ROOT
# --no-track: this is an autonomous loop; its reads must NOT train q-values (scope-primitive learning hygiene)
python3 resources/local-brain-search/search.py "[central question]" --limit 8 --mode spreading --no-track 2>/dev/null
python3 resources/local-brain-search/search.py "[leading hypothesis]" --limit 5 --mode spreading --no-track 2>/dev/null

Environment note (Trinity container): the bare python3 .../search.py form fails on hosts without system-level faiss (ModuleNotFoundError: No module named 'faiss') - e.g. the Trinity scheduled-run container. Use the wrapper ./resources/local-brain-search/run_search.sh "[query]" [same flags] instead; it routes through the running daemon (or a venv fallback) and honors BRAIN_READ_SCOPE (defaults fail-closed to core). Same flags apply (--mode static|spreading, --threshold, --no-track, --limit).

Mode fallback for narrow technical topics: Spreading activation propagates through the graph's highest-centrality hubs, so for narrow/technical questions (e.g. AI-architecture, distributed-systems topics) it can return the emotional/neuroscience mega-hubs ([[Dopamine]], [[Flow is a selfless state]], [[In Buddhism - Self is an Illusion]]) as top hits - noise, not evidence. If spreading results are dominated by off-topic hubs, re-run the same queries with --mode static (vector similarity), which surfaces the precise topical notes. Use whichever mode yields on-topic evidence; static is often better for technical clusters, spreading for cross-domain/synthesis topics.

Threshold note (post-2026-06-25 cosine correction): The corrected cosine_ip index lowered raw query→note similarity scores, so the default --threshold 0.5 now frequently returns "No results found" even for on-topic queries (abstract/technical questions especially). If a search comes back empty, re-run with --threshold 0.25 and shorter, keyword-style queries before concluding the KB has no evidence. Treat 0.25-0.35 hits as candidate evidence to read and judge, not as automatic relevance.

Read the top 3-4 most relevant notes in full. These are the evidence base for this iteration.

Step 5: Apply the Thinking Move

Execute the move based on what was determined in Step 3. For each move:

ACH Audit

• List all current competing hypotheses
• For each piece of KB evidence found: does it support, contradict, or not apply to each hypothesis?
• Build a simple diagnostic matrix (evidence × hypotheses)
• Eliminate any hypothesis that is contradicted by multiple pieces of evidence
• Update confidence scores for surviving hypotheses
• Output: Pruned hypothesis list with updated confidence levels

Bayesian Update

• State prior confidence in leading hypothesis (from last run)
• Identify new evidence from KB search
• Explicitly reason: "Given this evidence, should I be more or less confident? By how much?"
• Use rough likelihood ratios (e.g., "this evidence is 3x more likely if H1 is true than if H2 is true")
• Compute and record updated posterior confidence
• Output: New confidence scores with explicit reasoning for the change

Steelman Opposition

• State the current leading hypothesis clearly
• Construct the strongest possible argument AGAINST it - not a strawman
• Search KB explicitly for evidence supporting the opposing view
• Assess: does the steelman change the leading hypothesis, or does it survive?
• If the steelman reveals a genuine weakness, revise the hypothesis
• Output: Steel-manned opposition + response + any hypothesis revision

Cross-Domain Bridge

• Identify the core mechanism or pattern in the current question
• Run KB search from an unrelated domain (e.g., if topic is about AI, search neuroscience or Buddhism)
• Find analogous patterns or contradictory evidence from that domain
• Extract what the foreign domain implies about the current question
• Output: Cross-domain insight + revised framing of the question (if warranted)

Implication Check

• Take the current leading hypothesis as given (temporarily)
• Derive 3-5 concrete, testable implications that should be true if the hypothesis holds
• Search KB and reason about whether those implications are actually supported
• External probe (this move only — no other move uses WebSearch): If an implication concerns time-sensitive empirical state the KB cannot resolve, run up to 2 web searches per topic per run to test it. Restrict queries to reputable sources via site: filtering — peer-reviewed (arxiv.org, nature.com, science.org, pubmed.ncbi.nlm.nih.gov), established outlets (economist.com, ft.com, reuters.com, apnews.com, bloomberg.com), analytical (hbr.org, quantamagazine.org). (This is a subset of the canonical allowlist in resources/SOURCE-AUTHORITY.md — kept inline for in-query reliability; keep in sync with that file.) Crawler caveat: several of these domains (reuters.com, apnews.com, ft.com, economist.com) are not accessible to the WebSearch user-agent and an allowed_domains probe restricted to them returns a 400 error — wasting one of the 2 probes. Do NOT pass those as a hard allowed_domains filter; instead run the query unrestricted (or allow only the crawlable reputable domains: bloomberg.com, nature.com, arxiv.org, hbr.org, and analytical trackers like CSIS/RUSI/AEI/Jamestown) and judge source quality manually. Tag each external finding as [EXTERNAL: source — date] in the run output so KB-grounded vs. web-augmented evidence is auditable.
• If an implication is clearly false, that is disconfirming evidence for the hypothesis
• Output: Implications assessed as supported/unsupported/unknown + impact on hypothesis confidence

Assumption Audit

• List the 3-5 key assumptions the current reasoning rests on
• Rank by: (a) how load-bearing for the conclusion, and (b) how uncertain/shaky
• Focus scrutiny on high load-bearing + high uncertainty assumptions
• For the shakiest assumption: search KB for evidence bearing on it
• Output: Assumption vulnerability map + revised conclusion if key assumption weakens

Step 6: Write Updated Thinking File

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

### Run [N]: [Move Name] - [YYYY-MM-DD]

**KB Evidence Consulted:**
- [[Note A]] - [one line on relevance]
- [[Note B]] - [one line on relevance]

**Analysis:**
[2-4 paragraphs of actual reasoning from the move]

**Updated Hypotheses:**
| Hypothesis | Confidence | Change |
|-----------|------------|--------|
| H1: [statement] | X% | ↑/↓/→ from Y% |
| H2: [statement] | X% | ↑/↓/→ from Y% |

**Current Best Answer:** [1-2 sentences - the leading position after this run]

**Open Questions for Next Run:** [What remains unresolved or most worth probing]

Update the frontmatter: updated, updated_by, run_count, last_run.

Step 7: Check Convergence

A topic has converged when ANY of the following holds:

Standard convergence (all four must be true):

• run_count >= 4 (minimum 4 cycles)
• Same hypothesis has led for 3+ consecutive runs
• Confidence delta between last two runs < 5 percentage points
• No major open questions flagged in last run

Framework exhaustion (alternate path - triggers automatic convergence):

• run_count >= 10 AND
• The top two hypotheses have not changed relative order for 5+ consecutive runs AND
• Neither leading hypothesis's confidence has moved by more than 2pp across those 5 runs

When framework exhaustion is detected:

1. Apply one final Assumption Audit move on the framework itself (not the evidence). Ask: is the analytical framework (hypothesis structure, key distinctions) discriminating between meaningfully different world-states, or producing precision on an empirically-indistinct difference?
2. Note the assessment in the run output explicitly
3. Converge regardless of whether external "open questions" remain — those questions belong to a different topic (often: a parameter or timing question, not the structural question that was asked)

This guards against the "Confirmation Bias as Hyper-Precise Prior" failure mode: an analysis that produces ever-finer confidence scores on a distinction that the evidence has already shown to be analytically empty.

If converged (either path):

• Set status: converged in the thinking file frontmatter
• Append the canonical Final Synthesis block (see below)
• Add note: **CONVERGED** - Ready for crystallization via /manage-thinking-topics crystallize [slug]
• Update registry status to converged

If not converged: continue to next topic.

Canonical Final Synthesis block (appended once when status flips to converged):

---

## Final Synthesis

**Conclusion:** [1-2 paragraphs in plain prose - the substantive answer]

**Confidence:** X% on H[N] (and any co-leading hypotheses)

**Key evidence:**
- [[Note A]] - [why it mattered]
- [[Note B]] - [why it mattered]
- [[Note C]] - [why it mattered]

**Convergence path:** [Standard | Framework exhaustion] after [N] runs

**Remaining uncertainty:** [1 paragraph - what's known to be unknown; what could change this]

**Related topics:** [[other-converged-slug]] (if causally connected)

This block is the authoritative source for crystallization. Crystallize mode reads this section first, falls back to scanning the last run's "Current Best Answer" only if absent (legacy topics).

Step 8: Update Registry

Update Brain/05-Meta/Thinking/THINKING-REGISTRY.md:

• Increment run count for each processed topic

• Update last_run date

• Update status (active → converged where applicable)

• On convergence, MOVE the topic's row from Active Topics to Converged (Ready for Crystallization) using this exact row shape:

  | [slug] | [runs] | [one-sentence conclusion, ≤200 chars] → [[slug]] | [YYYY-MM-DD] |
  

Registry size discipline (hard rules — the registry is an index, not a journal):

1. Registry frontmatter may contain ONLY: created, updated, created_by, updated_by, agent_version, and a single last_incubation_run field of AT MOST one line (~200 chars): timestamp + topic(s) + move + resulting status. Example: last_incubation_run: 2026-07-03 07:00 UTC — agent-zero-trust-primitive Run 1 ACH Audit, still active
2. OVERWRITE last_incubation_run in place each run. NEVER append last_incubation_run_prior_* or any other run-history keys.
3. NEVER write analytical narrative into the registry — full reasoning already lives in the topic file (Step 6) and the session changelog (Step 9). Duplicating it in the registry made it unreadable (~100k tokens) and taxes every skill that loads it.
4. Converged-table conclusion cells stay at ONE sentence; the authoritative synthesis is the topic file's ## Final Synthesis block, which crystallize mode reads directly.
5. If legacy oversized fields or rows exist in the registry, do NOT imitate them — follow this format.

Step 9: Write Session Changelog

Write to Brain/05-Meta/Changelogs/CHANGELOG - Incubation Loop YYYY-MM-DD.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
---

# Incubation Loop Session: YYYY-MM-DD

## Topics Processed

| Topic | Move Applied | Confidence Shift | Status |
|-------|-------------|-----------------|--------|
| [slug] | [move name] | [leading H]: Y% → Z% | active/converged |

## Notable Shifts
[Any hypothesis revisions, convergences, or surprising KB evidence]

## Converged Topics (Ready for Crystallization)
[List any topics that reached convergence this run]

---

Seeding a New Topic

To add a topic to the loop, 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
last_move: null            # optional - written by loop after each run
convergence_count: 0       # optional - written by loop; tracks consecutive runs meeting convergence criteria
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]

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

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

Then add to THINKING-REGISTRY.md:

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

---

Crystallization (Manual)

When a topic reaches status: converged, the human should:

1. Read the full thinking file
2. Run /synthesize-insights [topic-slug] to graduate conclusions to a permanent note
3. Archive thinking file: set status: crystallized
4. The permanent note becomes part of the KB for future searches

Do not automate crystallization - judgment calls about what conclusions deserve permanence belong to the human.

---

Error Handling

Error	Recovery
Registry missing	Create empty registry, log, exit cleanly
No active topics	Run Starvation Floor (Step 1a) - invoke /domain-watch to self-seed a topic from watched domains; only log and exit cleanly if it too produces nothing
KB search returns empty	Try alternate search terms from hypothesis text; if still empty, note in run log and skip KB-grounding for this move
Thinking file missing for active topic	Log warning in registry, skip topic
Run exceeds 45 minutes	Process topics in order; skip remaining, log which were skipped

---

Completion Checklist

• Registry read - active topics identified
• Each active topic: KB searched with relevant queries
• Each active topic: correct move in rotation applied
• Each active topic: thinking file updated with reasoning + confidence scores
• Convergence checked for each topic
• Registry updated (run counts, dates, status changes)
• Session changelog written to Brain/05-Meta/Changelogs/

---

Self-Improvement

After completing this skill's primary task, consider tactical improvements:

• Review execution: Were there friction points, unclear steps, or inefficiencies?
• Identify improvements: Could the thinking moves, convergence criteria, or file structure be sharper?
• Scope check: Only execution changes - NOT changes to the ACH/Bayesian/dialectic framework itself
• Apply improvement (if identified):
  • Edit this SKILL.md with the specific improvement
  • Keep changes minimal and focused
• Version control:
  • git add .claude/skills/incubation-loop/SKILL.md
  • git commit -m "refactor(incubation-loop): <brief improvement>"