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":"1.22","updated":"2026-07-18T00:00:00.000Z","changelog":["1.22: Assumption Audit trigger-definition rule — when a committed fork/leg has committed EFFECTS but its TRIGGER lacks a wire-class definition (e.g. a 'did not convene' day-close test), the audit's product is the committed trigger definition: a positive-wire path (fires at its arrival, stale-guards binding), a silence-based path gated to day-close PLUS one news-cycle allowance (bare day-close firing barred — the observability filter is readout lag, not event absence), and a provisionality/unwind rider on any silence-based firing (a later wire proving the event occurred supersedes-and-replaces in full — unwind, not booking-audit, because the firing was made on certified absence) (Run 30 hold-state pattern)","1.21: Implication Check sensor-validity probes — a probe may target the institutional validity of a committed observable surface (does the named microphone/institution currently sit/publish?), exempt from the injection-first bar because scans observe wires, not institutional status; a validity finding books as framework structure (sensor re-scope, ladder compression, named restoration re-open), never event weights, and silence from an institutionally-dark surface is voided as color (Run 5 Kuwait-Assembly-suspension pattern)","1.20: ACH matrix-integrity pass codified — when the rotation returns ACH to a topic whose posterior has not moved since the standing matrix was built, run an integrity pass (elimination check / tether check / new-evidence placement with an explicit diagnosticity verdict) instead of a rebuild; never eliminate a hypothesis inside the pre-arrival window of its own committed discriminator (Run 13/25 pattern)","1.19: Assumption Audit attribution-vs-booking separation — when an audit challenges the interpretive basis of a committed leg (choice vs inability, deliberate vs stray) but the leg's hypothesis is scoped such that the outcome disconfirms it under ALL readings (e.g. window-scoped), the leg stands at full weight and no booking-audit opens; the challenge's real content books as attribution structure on DOWNSTREAM machinery (re-arm/decay schedules, successor readings) with a named wire-class re-open condition (Run 24 choice-revelation pattern)","1.18: Implication Check decomposition pre-commitment rule — when a leading hypothesis's implications accrue asymmetrically across separable components (e.g. propensity vs location vs timing), commit the decomposition AND the post-arrival reading of the relevant committed leg BEFORE the discriminator fires (color, never weights), foreclosing both post-hoc escapes: inflating a leg-firing into full refutation, or discounting it because the disconfirmed component 'moved elsewhere' (Run 23 displacement pattern)","1.17: Cross-Domain Bridge pre-fire annotation rule — when a bridge's insight concerns a pre-registered event that has NOT yet fired, convert it into a pre-committed interpretive annotation on that event's committed legs (narrative color, never weights), fixed BEFORE the outcome is known - so arrival narratives cannot inflate or half-promote a leg after the fact (Run 22 announcable-not-package pattern)","1.16: Steelman Opposition survivorship rule — when a steelman of a STANDING RULE (a binding, guard, or booking convention, rather than a hypothesis) fails, do not leave the rule defended-by-argument only: arm an explicit falsifier condition (a named future observable that would re-open the question as a v1.10 booking-audit candidate at its own arrival, color instrument never booking), converting the surviving rule into an instrumented commitment (Run 21 fear-family dissociation-falsifier pattern)","1.15: Partial-trigger rule (v1.7) adapted to NAMED FLAGS — a flag carries no committed bands, so a partial execution has no evidential component to book at a band edge; the epistemic component books in full as a RE-ARM of the flag's remainder with a sharpened class test, and repeated same-class executions accrue grade within their stream, never fresh spread (Run 19 Ev-W pattern)","1.14: Step 7 exhaustion-window exclusion extended to CONTEXT-CLASS arrivals — a run where an arrival landed but booked nothing because no committed leg binds it at all (context-only streams like P5, rhetoric-class items) is stasis by discipline, same as empty/gated windows, and does not count toward the 5-run window","1.13: Bayesian empty-window consolidation is repeatable — a previously consolidated ledger that accretes amendments across subsequent runs is re-consolidated into a superseding version with per-cell run citations (Run 14 Ledger v2 pattern)","1.12: Step 7 exhaustion-window exclusion extended to GATED-arrival zero-drift runs — a run where an arrival landed but booked nothing because its booking authority sits at a committed future read-point is stasis by discipline, same as an empty window, and does not count toward the 5-run window","1.11: Framework-exhaustion sharpening (Step 7) — empty-arrival-window zero-drift runs do not count toward the 5-run exhaustion window (stasis by discipline is not stasis by exhaustion), and the step-1 framework audit is determinative: pending pre-registered discriminators inside the topic's committed window DEFER the trigger rather than converging past them","1.10: Guard-interaction rule (Step 5) — pre-registered ratio/laundering guards bind EVENT bookings; an analytical move (steelman, assumption audit) may revise a prior booking only via an explicit booking-audit that either preserves the guarded ratio (joint deflation/inflation) or documents why the revision is not the laundering the guard polices","1.9: Injection adjudication precedence (Step 5, all moves) — un-priced arrivals in a fresh domain-watch injection (structural-break candidates, actor inversions, ordered adjudication lists) are adjudicated and booked as argued arrivals BEFORE the rotation move runs; the move then operates on the post-adjudication posterior","1.8: Successor-topic pre-registration rule — inherited event labels do NOT inherit committed effects; a successor topic must re-derive every LR/band against its OWN hypothesis field, and cross-hypothesis couplings discovered by earlier moves (e.g. an option-value asymmetry) must be committed INTO the table so a future arrival cannot be post-hoc read as moving only one hypothesis","1.7: Partial-trigger adjudication rule — when a pre-registered event fires partially, book the epistemic component of its committed effect in full, the evidential component at the lower band edge, mark it PARTIALLY CONSUMED with an explicit re-fire condition, and record all movement as a P-table event (not framework drift)","1.6: Bayesian empty-window rule extended — when pre-registration is already complete, consolidate scattered P-table commitments into one standing reference (event/state/effect); slow-accrual evidence books at natural granularity, never per-slot","1.5: Step 9 run-slot clock caveat — take the changelog HHMM suffix from the Execution Context timestamp, not the container's `date` (which can lag hours and collide with an existing file)","1.4: Bayesian Update empty-arrival-window rule — posterior = prior with zero drift when nothing new arrived; spend the run pre-registering LRs/posterior bands for named watch events instead of manufacturing movement","1.3: Implication Check injection-first rule — score against a fresh (<24h) domain-watch evidence injection instead of firing duplicate web probes; bank the probe budget","1.2: Step 9 changelog filename now carries the UTC run-slot suffix (YYYY-MM-DD-HHMM) — the loop runs multiple times per day and date-only names collide","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.