coding-agent-exercise

Create or update a main Stata do-file (main.do) running all other scripts. Use for: setting up main.do; or updating main.do when a new do-file is created or renamed.

Human-in-the-loop decision policy for World Bank-style development research coding agents. Use when an agent works on Lumora training data, Stata do-files, survey/payment data, field quality checks, or slide updates and must decide whether to ask the human, use a default, document an assumption, or proceed.

Responsible-language and claim-discipline skill for development research and operations. Use when describing vulnerable households, survey staff, payment providers, field quality flags, welfare changes, targeting results, or policy/operational implications from synthetic or real World Bank-style data.

High-frequency field data quality skill for World Bank survey and operations workflows. Use when reviewing Lumora follow-up surveys, survey logs, payment records, back-checks, field visits, enumerator effects, completeness, distribution, consistency, or operational follow-up priorities.

Reproducible development microdata analysis skill for Stata. Use when cleaning Lumora household, follow-up, payment, survey log, field visit, or back-check data; building processed datasets; validating IDs and merges; creating Stata tables/figures; or adapting World Bank/DIME-style analysis scripts.

World Bank-style Stata do-file and analysis-script reader. Use when inspecting `.do`, Python, or R scripts to understand generated tables, figures, samples, indicator definitions, merge logic, hardcoded numbers, and the provenance of claims before updating a deck or briefing.

Programmatic PowerPoint update skill for World Bank-style project decks. Use when updating an existing baseline PPTX with new generated Stata figures, tables, output provenance, slide-source mapping, restrained institutional styling, and an update log.