mechanism-skills

mechanism-skills is a skill for Claude Code from zjunlp/Mechanist. It costs 214 tokens per session (6,965 once invoked), scanned A, original, MIT.

A routing guide for methods that study how a neural network produces behavior internally. Mechanistic interpretability means examining components such as layers, attention heads, neurons, or learned features and testing their influence.

In plain words
What is it for?
Use it for research involving causal interventions, component influence, model mechanisms, vision explanations, or evidence for claims about a model’s internal behavior.
Why use it?
It helps select a suitable internal-analysis method when external accuracy alone cannot show which part of a model caused a result.

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 for research involving causal interventions, component influence, model mechanisms, vision explanations, or evidence for claims about a model’s internal behavior.

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

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

Your own site · 80×15
<a href="https://agentmods.dev/skills/zjunlp/mechanist/mechanism-skills"><img src="https://agentmods.dev/badge/skills/zjunlp/mechanist/mechanism-skills.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 214 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 6,965 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
SkillSpector: 1 finding, up to low

These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →

  • low Excessive Agency · line 41
    Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.
    Fix: Limit the skill's scope to its documented purpose. Remove instructions that enable the agent to perform actions outside its stated functionality.
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 $0.00214 $0.06965
Opus 5 $0.00107 $0.03483
Sonnet 5 $0.00043 $0.01393
Haiku 4.5 $0.00021 $0.00696

Measured 11d ago against content hash 7cc9cbaec01d, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-11, from the pricing page.

Security

Grade A, and why

mechanism-skills 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 11d ago.

The scan reads SKILL.md. This mod also ships 22 executable files (causal-attribution/ablation/scripts/intervention_experiments.py, causal-attribution/ablation/scripts/model_analysis.py, causal-attribution/attribution-patching/scripts/ioi_dataset_usage.py, …), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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-skills/SKILL.md · 260 lines

How it starts

The opening of the file, as written. The whole thing — 260 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Mechanism Skills

A curated set of eleven method families for analyzing the internal mechanisms of neural networks. Use this document as the routing entry point: read it first, pick the family that matches the question, then follow the link into the sub-skill for submethod detail, demos, and reference implementations.

When to Use

Consult these skills whenever the research question is about how the model computes internally, not just how it performs externally. Typical triggers:

  • A reviewer asks for mechanistic evidence behind a claim
  • A claim attributes a behavior to a specific component (layer, head, neuron, feature, weight)
  • An experiment needs to isolate which internal object drives a behavior
  • A method swap is under consideration and the candidate pool should include mechanism-level tools (e.g., swapping a magnitude-based screen for a causal-intervention method)
  • A vision or vision-language model needs concept-level explanations rather than raw saliency

If the research question is purely behavioral (accuracy, robustness, calibration) and does not depend on internal structure, these skills are not the right tool — continue with the standard experiment skills.

Loading Protocol (Mandatory)

Skills cascade strictly top-down: this routing file → family SKILL.md → submethod SKILL.md. Each level adds detail the level above only sketches. Never act on a summary from a higher level alone.

Hard requirements — these are not suggestions:

  1. Pick a family based on the research question (use the Selecting a Method table). The summaries below are previews only — they are intentionally insufficient to plan or run work.
  2. Once a family is selected, you MUST load <family>/SKILL.md in full before any further reasoning, recommendation, or code in that family. Assumed objects, exact signal definitions, scope conditions, composition rules across submethods, and known failure modes live in the family file, not here.
  3. Once a submethod is selected, you MUST load <family>/<submethod>/SKILL.md in full before running, citing, choosing hyperparameters for, or recommending that submethod. Runnable scripts, API conventions, defaults, and submethod-specific gotchas live there.
  4. Do not skip levels. Going from this file directly to a submethod without reading the family SKILL.md is forbidden, because cross-submethod composition (screen-then-verify pairings, shared metrics, layer/position conventions) is defined at the family level.
  5. Re-load on switch. If the analysis pivots to a different family or submethod, repeat the cascade for the new branch — do not carry assumptions across families.

Read the full file on GitHub · 260 lines

Files

What ships with it

60 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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. 11d ago First seen · 260 lines · 214 tokens per session scan A 7cc9cbaec01d

Subscribe to this mod's changes

mechanism-skills is a skill published in the GitHub repository zjunlp/Mechanist (74 stars, last pushed 15d ago), licensed MIT. It adds 214 tokens to every session and 6,965 once invoked, about $0.0011 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.