Edward Cheng-I Wu

A research workflow with decisions left to the researcher

Academic Research Skills divides work into research, writing, review, revision, and finalization. Its pipeline asks the researcher to confirm stages and keeps artifacts for later inspection. Integrity checks address citations, data, and logical consistency, while a revision traceability matrix connects reviewer concerns to claimed changes. The project treats model-generated criticism as something to examine rather than an automatic editorial verdict. Its documentation distinguishes procedural instructions from guarantees that a model will follow them.[2]

Adapting the workflow to Codex

ARS-Codex packages the research workflow as a single Codex-native skill with a router and optional runtime adapter. It vendors an identified upstream version and keeps packaging and runtime behavior separate from the original distribution. Users can request a focused review, research-question discussion, or a staged pipeline. The README recommends starting the pipeline with an intake checkpoint instead of asking it to finish an entire paper silently.[3]

Critical Thinking for Humans turns the assistant into a coach. Drill mode asks users to identify gaps between evidence and conclusions; Scene mode examines reasoning without pretending that value judgments have a single answer key. The repository also offers exploration and detective exercises. Its browser Casebook contains fixed bilingual excerpts, while the CLI and chat skill support broader coaching sessions. The distinction matters for someone choosing between a quick demonstration and an ongoing practice workflow.[4]