mechanism-explore

mechanism-explore is a skill for Claude Code from zjunlp/Mechanist. It costs 164 tokens per session (2,073 once invoked), scanned A, original, MIT.

A research-planning guide for finding the internal cause of a known model behavior. It chooses strategic directions for studying where the behavior is computed, whether a component causes it, how it formed, and whether the result is trustworthy.

In plain words
What is it for?
Use it after a validated behavior has been identified to plan causal investigations into neural-network or multimodal-model mechanisms.
Why use it?
It separates observing that a model behaves a certain way from proving why it happens. That keeps explanations tied to experiments that can test cause and effect.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin.

Part of the mechanist plugin — 54 skills, 4 agents shipped together

Good fit Use it after a validated behavior has been identified to plan causal investigations into neural-network or multimodal-model mechanisms.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/zjunlp/mechanist/mechanism-explore
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent
npx skills add zjunlp/Mechanist --skill mechanism-explore
Clone the repo
git clone --depth 1 https://github.com/zjunlp/Mechanist

Made for: Claude Code.

Or install mechanist, the plugin that ships this one along with the rest of its 54 skills, 4 agents.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for mechanism-explore

README.md
[![agentmods](https://agentmods.dev/badge/skills/zjunlp/mechanist/mechanism-explore/github.svg)](https://agentmods.dev/skills/zjunlp/mechanist/mechanism-explore)
Your own site
<a href="https://agentmods.dev/skills/zjunlp/mechanist/mechanism-explore"><img src="https://agentmods.dev/badge/skills/zjunlp/mechanist/mechanism-explore/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for mechanism-explore

Your own site · 80×15
<a href="https://agentmods.dev/skills/zjunlp/mechanist/mechanism-explore"><img src="https://agentmods.dev/badge/skills/zjunlp/mechanist/mechanism-explore.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 164 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,073 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe. Third-party audits
  • NVIDIA SkillSpector pass 7 Sept 2026
How audits are shown
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00164 $0.02073
Opus 5 $0.00082 $0.01037
Sonnet 5 $0.00033 $0.00415
Haiku 4.5 $0.00016 $0.00207

Measured 9d ago against content hash ecb874f9e41a, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

mechanism-explore scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/mechanism-explore/SKILL.md · 76 lines

How it starts

The opening of the file, as written. The whole thing — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Mechanism — Explore

The explanation half of the loop. Given a validated phenomenon, this skill is the macro-level plan for finding the mechanism behind it. It decides which strategic directions to pursue, and in what order — the families in /mechanism-skills execute the chosen directions.

A mechanism claim is causal: "component X is responsible for behavior B" means intervening on X changes B in the predicted, specific way. Locating X is necessary but not sufficient — only intervention earns the word "mechanism."

When to Use

A phenomenon is in hand — whether already established/known or handed off from /mechanism-behavior-discovery — and the question is now where it is computed, whether that component causes it, whether it can be tuned for use, how it formed, what it means, or whether the model's decision is trustworthy. Do not use this skill to find a phenomenon (that is /mechanism-behavior-discovery), nor as a substitute for the chosen family's method file in /mechanism-skills.

The Six Research Directions

Six parallel directions for explaining a model. They are coordinate — each answers a different question and stands on its own — and they also chain into strategies (see below). Location is typically the natural entry point for the others, since most directions act on a component you have first located.

1. Location — where the behavior is computed

At inference time, find which internal function component for the behavior: a layer, a neuron (or head), a circuit, or a featur/activation direction. Use cheap correlational/attribution methods (probing, vocabulary projection, magnitude, attribution, circuit discovery, dictionary learning). Output: a ranked shortlist of candidates. This is correlational — a located component is a hypothesis, not yet the cause.

2. Causal Intervention — whether the component causes the behavior

Intervene on the located component and check the target behavior moves as predicted (amplify → behavior strengthens, ablate → behavior gone). Tools: ablation, activation patching (sufficiency / localization), steering (dose-response on a represented quantity). Always report sign, magnitude / dose-response, and specificity (a matched control component does nothing; off-target behavior intact). This is what promotes located to mechanism.

Read the full file on GitHub · 76 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 9d ago First seen · 76 lines · 164 tokens per session scan A ecb874f9e41a

Subscribe to this mod's changes

mechanism-explore is a skill published in the GitHub repository zjunlp/Mechanist (74 stars, last pushed 13d ago), licensed MIT. It adds 164 tokens to every session and 2,073 once invoked, about $0.0008 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories