This repository was archived by the owner on Aug 8, 2026. It is now read-only.
Record the two-register split and the cross-repo citation rule - #78
Merged
Conversation
npf grew its own handoffs register on 2026-08-05, so CLAUDE.md's handoff guidance was describing a world with one. Three things a session now needs and could not get from this file. The split is by SUBJECT, not by where the work runs, which is the part that catches people. A work order about this package lives here even when it executes in npf, and that is the normal case rather than the exception: the 10 pattern is authored here, executed there, verdict written there. Only work about npf's book goes to npf's register, because this package ships no strategy and must not read as commissioning strategy research. A moved handoff leaves a pointer row and no copy. Duplicating one rebuilds the hazard 2026-08-03-index-share.md records, where a stale OPEN, unclaimed row sat on top of finished analysis and invited a second execution whose numbers would not have matched the first. A citation that crosses repos must carry its repo and file. A bare C7 resolves here, where tests/test_references.py checks it, and resolves nowhere from npf. The test scans this repo only, so moving a document out moves its citations beyond the reach of the thing that was checking them. That is not obvious from either end and it is how a checked citation becomes an unchecked one without anyone touching it. The lifecycle table's handoffs row also gains the pre-execution correction rule it was missing: a fix needed before execution that touches no threshold, statistic, universe or claim is recorded in the file as a dated note rather than applied silently, per 7.7 of the 10 prereg. That precedent has now been used twice and was written down nowhere a reader would find it. No status list is added here. docs/handoffs/README.md is the authoritative table and a second copy is the failure this document keeps catching elsewhere: it goes stale silently and reads as current. 566 passed, 93 skipped. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Follow-up to #77 and npf#192.
npfgrew its own handoffs register, soCLAUDE.md's handoff guidance was describing a world with one.Three things a session now needs and could not get from this file.
The split is by subject, not by where the work runs
That is the part that catches people:
npfnpf's booknpf/docs/handoffs/. This package ships no strategy and must not read as commissioning strategy researchA moved handoff leaves a pointer and no copy
Duplicating one rebuilds the hazard
2026-08-03-index-share.mdrecords: a staleOPEN, unclaimedrow sitting on top of finished analysis, inviting a second execution whose numbers would not have matched the first.A cross-repo citation must carry its repo and file
A bare
§C7resolves here, wheretests/test_references.pychecks it, and resolves nowhere fromnpf. That test scans this repo only, so moving a document out moves its citations beyond the reach of the thing that was checking them. That is not obvious from either end, and it is how a checked citation becomes an unchecked one without anyone touching it.Also
The lifecycle table's handoffs row gains the pre-execution correction rule it was missing: a fix needed before execution that touches no threshold, statistic, universe or claim is recorded in the file as a dated note rather than applied silently, per §7.7 of the §10 prereg. That precedent has now been used twice and was written down nowhere a reader would find it.
What is deliberately not here
No status list.
docs/handoffs/README.mdis the authoritative table, and a second copy is the failure this document keeps catching elsewhere: it goes stale silently and reads as current. The new text points at the register rather than restating it.566 passed, 93 skipped. Every path linked in the new text verified to exist.
🤖 Generated with Claude Code