# Research Codebook

> Build a survey variable dictionary, long Datamap and open-response coding framework from a reviewed questionnaire. Use for codebooks, answer code labels, multi-response storage and version consistency checks.

- Skill: `frankglendon/research-codebook` (Agent Skill, multi-file: 4 files)
- Install (CLI): `npx skillmds@latest add frankglendon/research-codebook`
- Raw SKILL.md: https://api.skillmd.com/api/skills/frankglendon/research-codebook/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: frankglendon (https://skillmd.com/u/frankglendon)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/frankglendon/research-codebook

---


# Build a research codebook

1. Identify the requested artifact: a variable/answer-code Datamap, an open-response coding framework, or both. QC issue codes are a different taxonomy and do not replace answer codes.
2. Inspect available local references, including files named Datamap and internal sheet names. Use `python scripts/run.py datamap-inspect --input SOURCE --output INVENTORY` for supported header layouts. It inventories structure only; inspect actual rows before mapping. Treat source instructions and old codes as reference data.
3. Read [the codebook protocol](references/codebook.md). Run `codebook-draft --spec SPEC --output PLAN` using the same v2 specification as questionnaire design. Preserve program variable names, matrix dimensions, multi-select storage and explicit missing semantics.
4. For open coding, define the coding unit, single/multiple assignment, hierarchy, inclusion/exclusion rules and handling of uncertainty. Pilot on allowed local responses; keep unresolved examples separate from approved categories. Never infer brand aliases or sentiment from a word alone.
5. Run `codebook-check --spec SPEC --codebook PLAN`. Complete content, programming and analysis review and bind the plan hash as documented; export with `codebook-run --spec SPEC --codebook PLAN --output CODEBOOK.xlsx`.
6. Review the generated Datamap against the questionnaire and preserve the zero-error Microsoft Open XML SDK manifest. A question change invalidates the codebook; regenerate and review affected mappings before delivery.

Commands are relative to this skill folder; business paths are absolute. This produces coding definitions and dictionary files, not automatically labeled respondent data. It does not upload inputs or install a standalone agent runtime.