replicate-paper

To replicate experiments from a research paper PDF, you need an active research session and a file path or URL. The skill parses the methodology, creates an experiment list, and routes the document through planning, execution, and interpretation phases to reproduce the results.

Yes, you can validate benchmark claims from a mechanistic interpretability paper. The skill extracts reported quantitative findings into sidecar context, runs the replication pipeline, and generates a report comparing your outcomes against the original paper's claims.

You can plan and run experiments based on a published methodology by providing the paper text or PDF. The skill orchestrates experiment planning, execution, and interpretation stages, enforcing a one-session-one-plan guardrail to maintain a structured research workflow.

Yes, an active research session is required to reproduce a paper's results. The skill enforces session-safe workflow conventions, including interactive approval gates for extracted methodology and one session per plan, ensuring your replication draft artifacts are properly managed.

The best way to compare reproduction results against reported quantitative findings is using paper-grounded comparison artifacts. The skill builds paper-specific sidecar context from the original text and generates a final report with a clear paper-versus-replication comparison.

Yes, you can extend a published causal model study with additional experiments using the extracted experimental skeleton. The skill extracts the original methodology and experiment list, allowing you to utilize the same framework for planning and running your extended studies.