Summary
The import-format guide references RANK_SPEC.md for the full rank-battery import/validation flow, but that document is not reachable through the documented mechanism (zwill guide show) and does not appear in zwill guide list.
Repro
zwill guide show import-format includes:
A rank battery is a set of per-item questions scored together … See RANK_SPEC.md for the full rank flow.
But:
zwill guide list
# -> only: agent-workflow, interpreting-results, import-format
zwill guide show RANK_SPEC
# {
# "status": "error",
# "errors": [{ "code": "not_found",
# "message": "Unknown zwill guide: RANK_SPEC.",
# "context": { "known_guides": ["agent-workflow","import-format","interpreting-results"] } }]
# }
Impact
The canonical rank-battery import shape (how _RANK / _GROUP columns map to rank_task_id, expected answer encoding, direction) is effectively undocumented. I had to infer it and rely on the auto-detector. (The auto-detector did work well — it grouped the 4 batteries and inferred 1_is_best — but the reference doc it points you to is a dead link.)
Suggested fixes
- Ship
RANK_SPEC as a guide show-able guide (add it to the bundled guides), or
- Replace the reference in
import-format with the actual content / a reachable link.
Summary
The
import-formatguide referencesRANK_SPEC.mdfor the full rank-battery import/validation flow, but that document is not reachable through the documented mechanism (zwill guide show) and does not appear inzwill guide list.Repro
zwill guide show import-formatincludes:But:
Impact
The canonical rank-battery import shape (how
_RANK/_GROUPcolumns map torank_task_id, expected answer encoding, direction) is effectively undocumented. I had to infer it and rely on the auto-detector. (The auto-detector did work well — it grouped the 4 batteries and inferred1_is_best— but the reference doc it points you to is a dead link.)Suggested fixes
RANK_SPECas aguide show-able guide (add it to the bundled guides), orimport-formatwith the actual content / a reachable link.