Files and change rule
| File | Role |
|---|---|
master/Analysis methods design.md |
This document. |
schema/schema.json |
Field vocabularies, requirement types, working document types, check library, stakes levels, and the rules for steps 1 to 8. Version 1.2.23. |
master/master_tagged.csv |
The catalogue: 236 rows, 52 columns, with result formats. |
master/master_base.csv, master/tags/tags_A.json to tags_F.json |
Sources of the CSV text and tags. |
master/specs/out_1.json to out_6.json |
Result formats. |
scripts/intake.py, select.py, sequence.py, validate_tags.py |
Process scripts for steps 1, 2 and 4, and the tag validator. |
master/test_jobs/, intake_test/, master/card_test/ |
Test briefs, intake inputs and the result format test. |
master/spot_check.md |
The independent check of 30 rows. |
analysis.zip |
The skill: SKILL.md, references/ (this document, the schema, the catalogue and the prompts) and scripts/. The site repository keeps a copy as data/analysis.zip, which the site offers for download. |
data/use-by-hand.md, data/limits.md, data/change-log.md, data/research-findings.md |
The text of the pages Use by hand, Limits, Change log and Source research, in the site repository. |
run2/ |
The complete run of 13 November 2026: inputs, plan, results, combined findings and answer. |
Change rule
A change to a field, a requirement, a step or a method is made in schema.json and in this document together. A change to a technique's text or tags is made in the source files and the CSV is rebuilt. When the skill changes, replace data/analysis.zip in the site repository with the new zip, so that the download matches the version in this document. Add a row to data/change-log.md for every new version. The build fails when the newest row does not match the version at the top of this document. When section 10 or the tested items change, update data/limits.md to match. After any change, run the validator (python3 -I scripts/validate_tags.py master/tags/tags_*.json), then run select.py and sequence.py on the test briefs and compare the counts with the last recorded run.