home / skills / plurigrid / asi / active-interleave

active-interleave skill

safe

This skill interleaves context from recent Claude/Amp threads to reduce uncertainty and accelerate current task insight.

npx playbooks add skill plurigrid/asi --skill active-interleave

Review the files below or copy the command above to add this skill to your agents.

Files (2)

SKILL.md

3.1 KB

---
name: active-interleave
description: Interleaves context from recently active Claude/Amp threads into current activity via random walk.
metadata:
  trit: 0
---

# Active Interleave Skill

Interleaves context from recently active Claude/Amp threads into current activity via random walk.

## bmorphism Contributions

> *"all is bidirectional"*
> — [@bmorphism](https://gist.github.com/bmorphism/ead83aec97dab7f581d49ddcb34a46d4), Play/Coplay gist

**Active Inference Pattern**: The interleave implements [Active Inference in String Diagrams](https://arxiv.org/abs/2308.00861) epistemic foraging — actively sampling from recent contexts to minimize uncertainty about the current task. Each random walk step is an epistemic action that gathers information.

**Play/Coplay Duality**: The interleave embodies bmorphism's bidirectional principle:
- **Play** (action): Query recent threads, walk the awareness graph
- **Coplay** (perception): Integrate fragments, update current context

**GF(3) Balanced Exploration**: The triadic structure (MINUS/ERGODIC/PLUS) ensures balanced exploration — validation filters (MINUS), random walk explores (ERGODIC), and colored emission generates (PLUS). Conservation Σ = 0 maintains coherence.

## Activation

Load when context from recent work would help current task.

## Usage

```bash
# Interleave from last 24 hours
bb ~/.claude/skills/active-interleave/active.bb

# Interleave from last N hours
bb ~/.claude/skills/active-interleave/active.bb --hours 6

# Query-focused interleave
bb ~/.claude/skills/active-interleave/active.bb --query "aptos blockchain"

# JSON output
bb ~/.claude/skills/active-interleave/active.bb --json
```

## Behavior

1. **MINUS (-1)**: Validate recency window, filter to active threads only
2. **ERGODIC (0)**: Random walk through recent sessions following awareness edges  
3. **PLUS (+1)**: Emit interleaved fragments with GF(3) coloring

## GF(3) Conservation

Each interleave batch maintains Σ trits ≡ 0 (mod 3).

## Integration

Call from current thread to surface relevant recent context:

```clojure
;; In any bb script
(require '[babashka.process :refer [shell]])
(def context (-> (shell {:out :string} "bb" (str (System/getProperty "user.home") 
                 "/.claude/skills/active-interleave/active.bb") "--json")
                 :out))
```

## Schema

Reads from `~/worldnet/cognition.duckdb`:
- `messages` - Content with timestamps
- `sessions` - Session metadata  
- `awareness_edges` - Play/coplay graph
- `temporal_index` - Time-ordered index



## Scientific Skill Interleaving

This skill connects to the K-Dense-AI/claude-scientific-skills ecosystem:

### Graph Theory
- **networkx** [○] via bicomodule
  - Universal graph hub

### Bibliography References

- `general`: 734 citations in bib.duckdb

## Cat# Integration

This skill maps to **Cat# = Comod(P)** as a bicomodule in the equipment structure:

```
Trit: 0 (ERGODIC)
Home: Prof
Poly Op: ⊗
Kan Role: Adj
Color: #26D826
```

### GF(3) Naturality

The skill participates in triads satisfying:
```
(-1) + (0) + (+1) ≡ 0 (mod 3)
```

This ensures compositional coherence in the Cat# equipment structure.

Overview

This skill interleaves context from recently active Claude/Amp threads into the current session using a random-walk awareness graph. It actively samples recent messages and emits concise context fragments balanced by a GF(3) triadic scheme. The result is relevant, temporally grounded context that can reduce uncertainty in ongoing tasks.

How this skill works

The skill reads recent messages, session metadata, and awareness edges from a local cognition store and validates recency with a MINUS filter. It then performs an ERGODIC random walk over the awareness graph to forage for informative fragments. Finally, a PLUS emitter colors and packages fragments into interleaved batches that conserve GF(3) trit sum ≡ 0, preserving compositional coherence.

When to use it

You are continuing work and want relevant fragments from recent threads.
You need quick, grounded context reminders from the last hours or days.
You want to reduce uncertainty by sampling related conversations before making decisions.
You are exploring connections between parallel sessions or emerging threads.
You want automated context synthesis for iterative scientific or technical drafting.

Best practices

Limit recency window to the most relevant timeframe (hours to a few days) to avoid noise.
Use targeted query mode when you need domain-focused interleaving (keywords or topics).
Review emitted fragments before incorporating them into authoritative output to ensure relevance and accuracy.
Combine interleaved context with your own notes rather than treating it as a primary source.
Monitor batch size to keep interleaving concise and cognitively manageable.

Example use cases

Resume a technical design discussion by surfacing recent decisions and open questions from related sessions.
Seed a draft with distilled points from adjacent experiments or threads to maintain continuity.
Explore hypotheses by linking observations across recent sessions during an epistemic foraging walk.
Quickly gather recent references and citation hints when assembling a literature or bibliography update.
Integrate fragments into a planning session to expose alternatives and prior constraints.

FAQ

How does the GF(3) coloring affect outputs?

GF(3) coloring tags fragments as MINUS (validation), ERGODIC (exploration), or PLUS (generation) and enforces that each batch balances to a net zero trit sum, keeping outputs coherent and compositional.

Can I restrict which sessions are considered?

Yes. Use a recency window, session filters, or a query string to narrow the graph walk to specific topics or time ranges.