mi-auditor
CommunityAudit MI claims with rigorous standards.
Education & Research#auditing#replication#mechanistic interpretability#statistical rigor#causal validity#R_V
AuthorAmitabhainArunachala
Version1.0.0
Installs0
System Documentation
What problem does it solve?
This Skill provides rigorous, multi-dimensional auditing of mechanistic interpretability research claims, ensuring scientific validity and reproducibility.
Core Features & Use Cases
- Multi-dimensional Auditing: Assesses claims based on statistical rigor, causal validity, cross-architecture replication, and literature positioning.
- Model Validation Tiers: Categorizes research findings into Ironclad (Tier 1), Discovery (Tier 2), and Problematic (Tier 3) based on validation status.
- Use Case: A researcher can submit their findings on a new interpretability technique, and the Skill will provide a comprehensive report detailing strengths, weaknesses, and necessary next steps for publication.
Quick Start
Use the mi-auditor skill to audit the claim that R_V contraction is causally linked to induction head formation.
Dependency Matrix
Required Modules
pypdfpdfplumberpdf2image
Components
scriptsreferences
💻 Claude Code Installation
Recommended: Let Claude install automatically. Simply copy and paste the text below to Claude Code.
Please help me install this Skill: Name: mi-auditor Download link: https://github.com/AmitabhainArunachala/clawd/archive/main.zip#mi-auditor Please download this .zip file, extract it, and install it in the .claude/skills/ directory.
Agent Skills Search Helper
Install a tiny helper to your Agent, search and equip skill from 223,000+ vetted skills library on demand.