菜单
Assess whether the research problem/behavior is important — its potential value and reach. Use when user says "impact check", "check impact", or wants to judge whether an idea/behavior matters before committing.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Representation and Parameter Analysis interprets and controls a model by directly manipulating its two kinds of internal objects — features (the hidden-state activations produced during a forward pass) and weights (the parameters of the target model). The analysis side relies on standard linear-algebra tools applied straight to these objects: techniques such as PCA, mean-difference, or simple linear classifiers are run over collected features (e.g., from contrastive prompt pairs) or over weight differences between a fine-tuned and a pre-trained checkpoint, in order to extract a small set of meaningful directions. The control side then operates directly through arithmetic on those objects: adding such a direction back into the features steers behavior at inference time, while adding, subtracting, or combining task-level weight differences edits the model in one shot — so that interpretability findings translate into behavioral control without any retraining.
Use this skill for feature-level steering of models — locating the internal feature that drives a target behavior, scoring and selecting it by its effect on the model's output, and directly amplifying or shrinking that feature's activation during generation to control behavior. Applies to features read from the model's own activations or from a Sparse Autoencoder (SAE); the bundled demo scripts happen to use an SAE, but the method does not require one.
Routing entry point for eleven families of mechanistic-interpretability methods that localize *which* internal object (layer, attention head, neuron, SAE feature, weight, or input feature) drives a model's behavior, *how influential* it is, and *what changes* when it is intervened on. Use this skill whenever the question is about a model's internal mechanism rather than its external metrics — for example, claims that a specific component is responsible for a behavior, mechanistic-evidence requests, circuit-discovery tasks, feature-attribution work (SHAP), or concept-level explanations of vision/VL models. The file lays out each family's premise, signal, cost, advantages, limitations, and how to compose them into a cheap-screen → causal-verify pipeline. Loading is hierarchical and mandatory: after picking a family from this routing file you MUST load that family's `SKILL.md`, and after picking a submethod you MUST load that submethod's `SKILL.md` — never act on the previews in this file alone.
Vocabulary Projection methods interpret internal model states by projecting them through the unembedding matrix to obtain a distribution over the vocabulary. The core idea is that the unembedding matrix, which maps the final hidden state to output logits, can serve as a universal decoder for intermediate states throughout the model. By inspecting the top-ranked tokens of the resulting distribution, researchers can directly read off the semantic content encoded in any internal object — such as a residual stream state or an attention head output — in terms of the model's output vocabulary.
Use when main results pass result-to-claim (claim_supported=yes or partial) and ablation studies are needed for paper submission. The external LLM reviewer (via llm-chat MCP) designs ablations from a reviewer's perspective, CC reviews feasibility and implements.
Search, download, and summarize academic papers from arXiv. Use when user says "search arxiv", "download paper", "fetch arxiv", "arxiv search", "get paper pdf", or wants to find and save papers from arXiv to the local paper library.
| name | impact-check |
| description | Assess whether the research problem/behavior is important — its potential value and reach. Use when user says "impact check", "check impact", or wants to judge whether an idea/behavior matters before committing. |
| argument-hint | ["idea-or-behavior-description"] |
| allowed-tools | WebSearch, WebFetch, Grep, Read, Glob, mcp__llm-chat__chat |
Check whether a proposed idea studies an important problem/behavior — one worth committing effort to: $ARGUMENTS
An idea is high-impact to the extent it does ANY of the following:
llm-chat — External LLM reviewer via llm-chat MCP (model defers to LLM_MODEL env). Always ask the external reviewer for strict, high-rigor feedback.Given an idea / behavior description, systematically assess its importance:
For EACH impact dimension below, judge how strongly the idea scores and gather supporting evidence (use WebSearch / WebFetch to check whether the problem is an open/active question, how much attention it gets, and where it would be used):
Call the external LLM reviewer via mcp__llm-chat__chat. Always ask for strict, high-rigor feedback.
Prompt should include:
Output a structured report:
## Impact Check Report
### Studied Problem/Behavior
[1-2 sentence description — the behavior/problem, not the method]
### Impact Dimensions
1. Important problem — HIGH/MEDIUM/LOW — [why]
2. Uptake / citation — HIGH/MEDIUM/LOW — [why]
3. Direction-shifting — HIGH/MEDIUM/LOW — [why]
4. Real-world reach — HIGH/MEDIUM/LOW — [why]
5. Phenomenon value — HIGH/MEDIUM/LOW — [why]
### Overall Impact Assessment
- Score: X/10
- Recommendation: PROCEED / PROCEED WITH CAUTION / DEPRIORITIZE
- Why it matters: [the strongest case, one sentence]
- "So what?" risk: [the strongest objection a reviewer would raise about importance]
- Who would care: [the audience that benefits most]
After each mcp__llm-chat__chat reviewer call, save the trace following shared-references/review-tracing.md. Write files directly to .mechanist/traces/<skill>/<date>_run<NN>/. Respect the --- trace: parameter (default: full).