audit-explanatory-compression

SkillDev tools

Lets your agent check whether a simple explanation is genuinely powerful or just relabeling observations in fancy words.

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 audit-explanatory-compression skill

About this skill

Test whether an elegant/simple explanation earns its compression by forbidding alternatives, subsuming independent facts, or making risky predictions, rather than merely relabeling observations with a compact vocabulary.

What this skill tells your AI

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

Purpose

Test whether an elegant explanation earns compression by excluding alternatives, subsuming independent facts, or making risky predictions.

Input contract

mode_contracts:
  earned-simplicity: &compression_audit_input
    required: [explanation, covered_facts, alternatives]
    optional: [predictions, evidence]
    constraints: [facts_and_alternatives_must_be_independently_enumerated]
  decorative-simplicity: *compression_audit_input
  risky-prediction: *compression_audit_input

Execution protocol

Do not perform called SOP operations inline; each loaded SOP owns its contract and thresholds.

  1. You MUST load skill classify-simplicity-evidence to classify the explanation's compression evidence.
  2. You MUST load skill test-risky-prediction to derive and test risky predictions.
  3. You MUST load skill construct-critique to attack the explanation. You MUST load skill score-object to score its remaining support. Deviation: omit prediction testing only when no nontrivial prediction can be derived, and mark the explanation non-discriminating.

Output contract

mode_contracts:
  earned-simplicity: &compression_audit_output
    produces: [forbidden_set, risky_predictions, accommodation_audit, deletion_test_result, elegance_verdict, earning_prediction]
    delta_fields: [findings, evidence_updates, uncertainties, decisions]
  decorative-simplicity: *compression_audit_output
  risky-prediction: *compression_audit_output

Thresholds and quality gates

  • Earned simplicity requires at least one independent fact compressed and one risky alternative-forbidding prediction, each evidence-linked.
  • Decorative simplicity is a failure when it merely renames observations.

Failure and counterexamples

Do not reward brevity alone. Mark weak when independent facts, exclusions, or predictions are absent.

Provenance map

  • resolved: elegance-trap-probe

Preserved source criteria ledger

sourcesource linekindsource criterion
v4 architecturenode desctextualDistinguish earned simplicity from decorative relabeling.

Context checkpoint / Delta notes

Append covered facts, alternatives, risky predictions, critique, and score rationale.

Mode branches

  • earned-simplicity: seek independent compression and exclusions.
  • decorative-simplicity: test for relabeling.
  • risky-prediction: prioritize prospective constraints.

Signals

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