Scaffold session-local analysis code and Hydra configs from spec markdown files.
npx skills add https://github.com/goodfire-ai/causalab --skill setup-analyses
Please help me install this Agent Skill. Skill: setup-analyses Source: https://github.com/goodfire-ai/causalab/tree/main/.claude/skills/setup-analyses Command: npx skills add https://github.com/goodfire-ai/causalab --skill setup-analyses
Setting up a new research analysis in Causalab is time-consuming and error-prone because it requires creating the right session-local package structure and Hydra config so the runner can execute it later.
${SESSION_DIR}/code/analyses/<name>/ (including main, init, README, and the spec file) and ${SESSION_DIR}/code/configs/analysis/<name>.yaml from one or more spec paths.cfg.experiment_root, and that layering rules are respected (analyses orchestrate methods and perform I/O via causalab.io.*).Use the setup-analyses skill to scaffold one analysis from an existing spec file at 'path/to/set_up_analysis.md' by running the command: /setup-analyses path/to/set_up_analysis.md
High-intent search queries and answers about installing and using this skill.
To scaffold a research analysis, provide a spec markdown file path to generate session-local code packages and Hydra configs simultaneously. The process creates analysis wrappers, init files, and YAML configurations that route outputs through the experiment root directory.
Hydra config wiring defines experiment parameters and output paths for research workflows. Scaffolding analysis code with Hydra ensures defaults live in configuration files rather than inline code, maintaining Causalab layering rules where analyses orchestrate methods and handle I/O separately.
Batch provisioning scaffolds multiple analysis specs sequentially from provided spec paths. In interactive mode, the runner processes each spec file with a single batch-level approval checkpoint, generating corresponding code directories and Hydra YAML configs for each analysis component.
Yes, scaffolding analysis code requires YAML frontmatter containing name and description fields in the SKILL.md file. The spec must also follow a 5-section structure to correctly generate both the session code package and the Hydra configuration file.
Analysis defaults must live in Hydra configs to enforce Causalab layering conventions and ensure output paths route through the experiment root. This separation prevents hardcoded paths in code, allowing the runner to execute analyses reproducibly across different experiment configurations.
Scaffolding session-local analyses is limited to Causalab environments requiring Hydra wiring and YAML frontmatter in specs. It does not support frameworks outside the Causalab layering system, and analyses must orchestrate methods while performing I/O exclusively through causalab.io modules.
eduardoabreu81Create a standardized SKILL.md structure with name and description frontmatter fields.
goodfire-aiExecute causal abstraction experiment plans into validated task scaffolding and runner YAML pipelines.
AutisticAFGenerate reusable contract test suites for Swift protocols.
goodfire-aiScaffold session-local interpretability method stubs from set_up_method specifications.