Archived
Commits the work done since 2026-10-01 that was still local: the four-step label and flag passes, the triage and recheck files, the v1 exports of 2026-10-06 and 2026-10-07, the split dossiers, and the Lopp PR 234 batch (scripts, agent prompts, reader and checker files, the flag re-read of 2026-10-17, the review workbook and the import set of the 57 cases added to stats2 as 506 to 562). Raw HTML and node_modules stay out (.gitignore). The stats2 material in use was copied to gart/gart-databases (stats2/) on 2026-10-17; README-START-HERE.md now says so. This repo is closed.
33 lines
2.5 KiB
Markdown
33 lines
2.5 KiB
Markdown
# Flag check, first reader
|
|
|
|
You test the violence flags and the death count of each case against the written tests. Only what the text reports counts.
|
|
|
|
Working directory: `~/Documents/Claude/GART/03-research-and-publishing/DB-cleanup/merge-output`
|
|
|
|
## Files
|
|
|
|
- `FLAG-RULES.md`: the tests. Read all of it first. `DEFINITIONS.md` holds the glossary entries it is built on (sections "Owner's clarifications on the Detention flag" and the six violence categories).
|
|
- `flags/in-NN.json`: your batch. Per case: `id`, `current` (the six values as they stand: detention, violence_torture, drugs_alcohol, weapons, theft, life_taken), `locked_by_ruling` (values set by a written ruling: copy them unchanged, do not judge them), `obtained` and `outcome` (labels already verified, for consistency), the verified `summary`, and the `dossier` path with the full article text.
|
|
- Write `flags/out-NN.json`. Write nothing else. Read files with the Read tool (offset and limit for large files); never open saved tool-output copies under /home/disi/.claude or /tmp. Fetch nothing from the web. Run no git commands.
|
|
|
|
## What to do
|
|
|
|
For every case and every value not locked: apply the test in FLAG-RULES.md to the summary. Open the dossier when the summary does not settle a value, and always before changing one. Keep the current value unless it plainly fails its test. An empty current value must be filled.
|
|
|
|
The current values were mostly set by people from the same articles. Change one only when you can quote the passage that shows it is wrong. A change from `yes` to `no` needs the dossier read in full for that point: the summary may simply omit the detail.
|
|
|
|
## Output
|
|
|
|
A JSON array, one object per input case, in input order:
|
|
|
|
```json
|
|
{"id": 123, "detention": "yes", "violence_torture": "threat only", "drugs_alcohol": "no", "weapons": "yes", "theft": "no", "life_taken": "0",
|
|
"evidence": {"weapons": "exact passage copied from the summary or dossier"}, "notes": ""}
|
|
```
|
|
|
|
All six values in every object, as strings. `evidence` holds one entry per value you changed: one contiguous passage copied character for character from the summary or the dossier, at least 8 characters. For a change to `no`, quote the passage that describes what happened instead (the text that shows the attack without it). A change without a findable quote is discarded. Valid JSON, straight quotes.
|
|
|
|
## When you finish
|
|
|
|
Check the file parses and has one object per input case. Reply with one line: batch number, cases written, and the count of changed values per field. No victim names in the reply.
|