Verification checks for the notebook interview¶
Generated by python scripts/check_interview_claims.py on 2026-09-18 from agent_skills_sample.db (read-only). Aggregate counts only. elicitation/notebook-interview.md cites these IDs.
| Check | Questions | What is checked | Result |
|---|---|---|---|
| V01 | NB-Q01 | Occurrences, representatives, distinct contents | 29,786 occurrences; 13,000 rows with dedup_primary = 1; 13,000 distinct file_sha |
| V02 | NB-Q01, NB-Q07a | Every distinct content has exactly one representative | 0 violations |
| V03 | NB-Q01 | Non-representative rows that carry content |
16 rows (documentation: representatives only); 16 of them identical to their representative's text |
| V04 | NB-Q02 | Filename case variants | SKILL.md 29,408, skill.md 349, Skill.md 24, SKILL.MD 5; non-SKILL.md names 378 (1.27%) |
| V05 | NB-Q02 | Front matter valid among representatives | valid 11,291; not valid 1,709 (86.9% valid) |
| V06 | NB-Q03 | Lowest-hash sampling rule | largest file_sha starts 01b3e7, at 0.6651% of the hash space; 13,000 of 1,877,981 is 0.6922% |
| V07 | NB-Q04, NB-Q05 | Bundled files with and without text | 47,829 rows; 22,278 with text; 8,783 directories; files without text: content_fetched = 1: 71, content_fetched = 2: 16,697 (859 over 100 KB); rows with a skipped_reason: 0 |
| V08 | NB-Q04 | Bundled-file rows whose (repo, path) is not a representative | 13 rows |
| V09 | NB-Q05, NB-Q05a | content_sha_ok among representatives |
1: 12,849, 2: 151 |
| V10 | NB-Q05 | composition_truncated and missing folder listings |
None: 10, 0: 11,235, 1: 1,755; mean sibling_count 8.5 when truncated vs 2.9; representatives without a folder listing: 10 |
| V11 | NB-Q06 | Commit history coverage by location (representatives) | canonical: 1,755 of 1,762; other: 442 of 3,822; skills-dir: 813 of 7,416. Dates without history_fetched: 0; history_fetched without a date: 0; history on non-representatives: 0 |
| V12 | NB-Q07a | Copies kept with each sampled content | 3,378 contents have 2 or more copies (2,854 in 2 or more repositories); most copies of one content: 188 |
| V13 | NB-Q07, NB-Q07a | Same-name families with more than one content | 542 front-matter names shared by 1,631 distinct contents |
| V14 | NB-Q12 | Email-like strings and masking markers in commit messages | 6,020 messages; 0 contain an email-like string; markers: <<redacted>> 2,570, <redacted> 304 |
| V15 | NB-Q12 | How co-author names in Co-authored-by trailers appear |
2,476 trailers: AI assistant named 2,333, fixed marker 130, bot login 8, other 5 |
| V16 | NB-Q12 | First-commit author values by account type | User length 16: 2,545, (empty) length 16: 436, Bot length 19: 18, Organization length 16: 4, Bot length 7: 3, Bot length 17: 1, Bot length 21: 1, Bot length 24: 1, Bot length 30: 1 |
| V17 | NB-Q13 | License of each representative's repository | (none) 6,992, MIT 3,789, Apache-2.0 1,044, NOASSERTION 742, AGPL-3.0 182; no usable license (none or NOASSERTION): 7,734 of 13,000 (59.5%) |
| V18 | NB-Q14 | has_scripts against P-01's script-extension list |
has_scripts 1 with a script file 1,326, without 0; has_scripts 0 with a script file 57, without 11,607 |