Scaffold session-local interpretability method stubs from set_up_method specifications.
npx skills add https://github.com/goodfire-ai/causalab --skill setup-methods
Please help me install this Agent Skill. Skill: setup-methods Source: https://github.com/goodfire-ai/causalab/tree/main/.claude/skills/setup-methods Command: npx skills add https://github.com/goodfire-ai/causalab --skill setup-methods
It removes the friction of creating new session-local interpretability “method” primitives by turning a structured method spec into runnable Python code with tests and documentation, following strict layering rules.
${SESSION_DIR}/code/methods/<name>/ with an importable module, a placeholder implementation, and a smoke test.Run /research-session first, then invoke setup-methods with one or more method spec paths, for example: “/setup-methods path/to/method_1/set_up_method.md path/to/method_2/set_up_method.md”.
High-intent search queries and answers about installing and using this skill.
Scaffolding reusable interpretability methods turns a structured set_up_method spec into runnable Python code with tests under a session directory. It generates an importable module, placeholder implementation, and smoke test while enforcing layering rules.
Generating Python method stubs with shape-and-dtype test scaffolds uses batch or interactive spec elicitation to produce in-memory, non-disk I/O method primitives. It writes code under the session directory with strict layering invariants and no hyperparameter defaults.
Batch scaffolding multiple interpretability method specs is supported by providing one or more spec paths to the tool. In interactive mode, it presents a consolidated batch summary so you can approve or edit specs before code generation begins.
Scaffolding interpretability method primitives enforces import restrictions by refusing forbidden imports from runner or analyses modules. This keeps methods pure with no disk I/O and no hyperparameter defaults in the generated code.
An active research session is required before scaffolding mechanistic interpretability methods because the tool writes generated code under the session directory. You must run the research session initialization first, then invoke the scaffolding with method spec paths.
Limitations of scaffolding session-local interpretability method primitives include strict enforcement of in-memory operations with no disk I/O and no hyperparameter defaults in code. Forbidden imports from runner or analyses modules are refused to maintain layering invariants.
infinitegameplayerInitialize AI interface sessions with continuity checks and approval-gated setup.
zli5460Train and analyze sparse autoencoders to interpret neural network activations.
goodfire-aiScaffold session-local analysis code and Hydra configs from spec markdown files.
SoSoBurnerRestrict file edits to a specified directory for the session.