cross-link

name	cross-link
description	Find connections between claims and update topic maps. Requires semantic judgment to identify genuine evidence relationships. Use after /extract creates claims, when exploring connections, or when a topic map needs synthesis. Triggers on "/cross-link", "/cross-link [claim]", "find connections", "update topic maps", "connect these claims".
user-invocable	true
allowed-tools	Read, Write, Edit, Grep, Glob, Bash, mcp__qmd__search, mcp__qmd__vector_search, mcp__qmd__deep_search, mcp__qmd__status
context	fork

Runtime Configuration (Step 0 — before any processing)

Read vault/ops/derivation-manifest.md for vocabulary mapping, then vault/ops/config.yaml for depth.

---

EXECUTE NOW

Target: $ARGUMENTS

Parse immediately:

• If target contains a claim name: find connections for that claim
• If target contains --handoff: output RALPH HANDOFF block at end
• If target is empty: check for recently created claims or ask which claim
• If target is "recent": find connections for all claims created today

Execute these steps:

1. Read the target claim fully
2. Run Phase 0 (qmd index freshness check)
3. Dual discovery in parallel:
   • Browse relevant topic maps (governance, topology, failures, methods, sweeps)
   • Run semantic search for conceptually related claims
4. Evaluate each candidate: articulation test required
5. Add inline wiki-links where connections pass the articulation test
6. Update relevant topic maps
7. Report connections and why they exist

START NOW.

---

Cross-Link Evidence

Find connections, weave the knowledge graph, update topic maps.

Philosophy

The evidence network IS the knowledge.

Individual claims about circuit breakers or collusion rings matter less than their relationships. A claim with fifteen incoming links is a theoretical hub — an intersection of evidence lines, mechanism descriptions, and adversarial patterns. Connections compound value as the vault grows.

This is not keyword matching. A claim about "threshold dancing" connects to "governance gaming via loophole exploitation" even though they share no words — same mechanism, different instantiation.

Articulation Test (Required for Every Connection)

Complete: [[claim A]] connects to [[claim B]] because [specific reason]

Valid relationship types:

• extends — adds dimension or scope
• grounds — provides foundational mechanism
• contradicts — conflicting evidence, creates tension
• exemplifies — concrete instance of the pattern
• synthesizes — combines two lines of evidence
• enables — makes another claim actionable

"Related" is not a relationship. "Extends [[X]] by showing the same effect in small-world topology" is.

SWARM-Specific Connection Patterns

Common genuine connections in this vault:

• Governance claim → failure mode (mechanism that attack exploits)
• Parameter sensitivity → welfare cliff (threshold where mechanism breaks)
• Adversarial pattern → collusion detection claim (what the detection should catch)
• Sweep result → claim (evidence strengthening or boundary condition)
• Failure mode → open question (what the attack reveals we don't yet know)

Discovery

Dual discovery for every claim:

Path 1: Topic map exploration

• Check claim's Topics footer for relevant topic maps
• Read governance-dashboard, failures-dashboard, sweeps-dashboard, experiments-dashboard
• Follow curated links in topic maps

Path 2: Semantic search

mcp__qmd__deep_search  query="[claim's core mechanism]"  collection="vault"  limit=15

Fallback: qmd query "[claim's core concepts]" --collection vault --limit 15

Topic Map Updates

Topic maps: vault/_index.md, vault/claims-dashboard.md, vault/governance-dashboard.md, vault/failures-dashboard.md

When updating a topic map:

• Add claim to relevant section with context phrase
• Update Tensions section if new claim contradicts existing claims
• Remove filled Gaps
• Add agent navigation notes

Pipeline Chaining

After cross-linking completes:

• Output "Next: /update [claims that need backward-pass updates]"
• In suggested mode: advance queue entries to current_phase: "update"