extract-causal-structure

SkillDev tools

Lets your agent pull causal claims, mechanism chains, and assumptions out of source material as a structured graph.

Available today. Use it from your connected AI after setup.

Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.

Then ask your AI: use the extract-causal-structure skill

About this skill

Extract causal structure from source material at requested granularity: individual causal claims, ordered mechanism chains, mediators, assumptions, and stated boundary conditions.

What this skill tells your AI

The instructions your AI receives, as published by yogsoth-ai/de-anthropocentric-research-engine in skills/extract-causal-structure/SKILL.md and read by ahel’s review.

Purpose

Extract causal claims, mechanism chains, mediators, assumptions, and boundary conditions at caller-specified granularity.

Input contract

required: [source_material, extraction_granularity, causal_schema]
optional: [theory_description, domain_ontology, evidence_links]
constraints: [preserve direction and scope; distinguish observed association from asserted causation; attach evidence]

Procedure

  1. Identify candidate cause, mediator, effect, conditions, and temporal order.
  2. Extract X->mediator->Y chains and direct X->Y claims at the requested granularity.
  3. Record assumptions, boundary conditions, and evidence for each edge.
  4. Assemble the causal graph and flag unsupported or ambiguous links.

If the extracted causal relations contain incompatible directions or effects under shared scope, consider detect-contradiction as the next tactic.

Output contract

produces: [causal_claims, mechanism_chains, causal_graph, boundary_conditions, evidence_links]
delta_fields: [findings, evidence_updates, hypothesis_updates, uncertainties]

Quality gates

  • Mechanism-extraction mode produces at least 1 chain per supplied theory and at least 2 chains in total where that source protocol applies.
  • Every edge has direction, scope, and evidence status; association is not upgraded to causation.
  • Biological strategy extraction preserves mechanism-level details of how function is achieved.

Parameterization

The caller must provide source material, theory or artifact schema, extraction granularity, causal edge vocabulary, boundary-condition fields, and evidence-link format.

Failure and counterexamples

Reject chains with missing direction, no stated mechanism, or evidence that supports only correlation while the output claims causation.

Provenance map

  • resolved: hypothesis-formation/mechanism-extraction
  • resolved: creative-ideation/biological-strategy-extraction
  • resolved: stress-test/causal-claim-extraction
  • intermediate: Pass3/extract-mechanism
  • intermediate: Pass3/extract-causal-claims

Preserved source criteria ledger

sourcephysical linekindsource criterion
hypothesis-formation/mechanism-extraction22numericProduce X->mediator->Y mechanism chains, at least 1 per theory and at least 2 total where applicable.

Signals

GitHub stars
501
Forks
41
Last commit
Sep 2026
Advanced
Catalog kind
skill
Key
extract-causal-structure
Source
github.com/yogsoth-ai/de-anthropocentric-research-engine