Diagram/text integration: diagrams replace reflection prints, trim whole-dump notebooks - #25
Merged
Merged
Conversation
…ompaction Addresses two findings: Phase 2's own diagram-survey table shows 14 of 16 placements were 'add' not 'replace', so dense text stayed in place next to every new diagram; every construction notebook prints its own assembled increment twice (once per fragment, once again as a 'reflection'). Scopes a harness-driven, two-phase fix (survey, then implementation) with explicit model/role discipline per DL-090's own lesson, folds in four already-identified cleanup items, and states the plan-sequencing requirement needed to stay durable across a context compaction.
…ect grep The original draft claimed Ch2/Ch3/Ch5/Ch8/Ch10 without checking each one; a repo-wide grep for the literal print(source) cell shows the real list is Ch2-03, Ch3-03, Ch4-03, Ch5-01, Ch6-03. Neither Ch8-01 nor Ch10-01 has this pattern at all. Caught before Phase A's own agents would have been sent out on a false premise.
Fully self-contained -- every agent's dispatch prompt is written out in full, not described, per writing-plans' own no-placeholder rule. States explicitly, on disk, that completing this plan triggers writing Phase B's own plan next, so the instruction survives a context compaction.
Nine parallel read-only chapter agents surveyed Ch1-Ch8 and Ch10 against the two target patterns. Orchestrator spot-checked 3 of 5 Pattern-2 proposals against the real committed SVGs (Ch2, Ch4, Ch6) -- all confirmed exact. Eight cross-chapter open questions surfaced, most load-bearing: every proposed fix in this survey contradicts tutorial-style-guide/SKILL.md's own "print TOASTER_INCREMENT as the reflection" rule, which needs one centralized skill-editor-gated fix before Phase B, not independent rediscovery per chapter. Also corrects a factual error found in decisions/diagram-survey.md (ch05's claimed "triple-dump" does not exist in the real files). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
…conflicts Both the construction-zone reflection print and the judgment-record tag reprint are ruled to be the same defect this redesign already targets, not sanctioned exceptions. Both edits trip skill-editor's own multi-archetype escalation row regardless of the ACE's ruling on the substance, so both entries stay PENDING until Z confirms the exact replacement wording. No skill file has been touched. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Z-confirmed wording. Applies Z's already-approved Decisions 1-2 (diagram/ text integration redesign) to the construction-zone reflection print and the judgment-record tag-anchor print, both of which this skill text had required and which every Phase A survey proposal now contradicts.
- decisions/log.md: DL-092 and DL-093 bodies still said "no edit has been made" / "PENDING Z's wording confirmation" after their headers were flipped to COMPLETE by fbdbc1d. Both now say the edit was made, citing fbdbc1d. - toaster-recipe/SKILL.md: the Pattern-B seam-cell sentence still pointed at "the TOASTER_INCREMENT string" as what the reader sees, which is stale now that TOASTER_INCREMENT is assigned, not printed. Points instead at the already-printed fragment variables and the diagram/confirmation query that fills the reflection role.
…n-not-print (DL-092/DL-093)
…pter contracts Fully self-contained -- every chapter task's acceptance criteria quote the Phase A survey's own literal cell content verbatim, plus the two DL-093-resolved judgment-tag fixes the survey itself held back as open questions. Uses this project's own CONTRACT.md-per-worktree harness (builder sonnet / reviewer opus), not writing-plans' generic two options, per spec Decision 5.
…emoval on 01-model-navigation; index.md heading sync
…n 03-completeness-check
…x.md separator fix
… confirmation added to 01-calc-energy
…ant-def to 01-assert-constraint-def; index.md separator fix
… full removal on 01-model-navigation; index.md heading sync
… trim on 03-completeness-check
…n 03-judgment-context
…n 03-threshold-judgment
…-invariant-def to 01-assert-constraint-def; index.md separator fix
….find() confirmation added to 01-calc-energy
… trim on 03-judgment-context
…n 03-stopping-judgment; index.md heading sync
… trim on 03-threshold-judgment
… trim on 03-stopping-judgment; 02-second-level Phase A coverage gap closed; index.md heading sync
…ation escalation closed (premise was false, no fix needed)
… DL-096 corrects DL-095 (2 Ch3 seam cells really are stale, follow-up contract dispatched)
…assembled-block print
…moved assembled-block print (DL-096)
…sting action-flow diagram)
…ing action-flow diagram (DL-094, option C)
… existing action-flow diagram (DL-094)
…plyHeat grounding fix
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
…p (analysis only, no implementation)
… once where made, query/retrieve in later chapters
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 join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Summary
TOASTER_INCREMENTand printed it a second time right after printing each fragment individually) across Ch1-Ch8 and Ch10. The chapter's own diagram, or a shortmodel.find()/model.query()confirmation where no diagram exists, now occupies that seam slot instead.print()the entire cumulative model file, to only the text their chapter's own diagram can't show. Ch5-01's residual is "none" — its diagram already covered everything structurally.chapters/ch08-checking/01-invariant-def.ipynbto01-assert-constraint-def.ipynb(its heading was already corrected to "assert constraint" in an earlier pass; the filename hadn't caught up), updating all 6 live references.index.mdheading case/separator drift across Ch5, Ch6, Ch8, Ch10.tutorial-style-guide,toaster-recipe, andtoaster-review-protocolskills soTOASTER_INCREMENTis documented as assigned-not-printed, including for judgment-record tag anchors — closing the exact contradiction every notebook fix above was otherwise violating.ch06/02-second-level.ipynbhad its own un-surveyed reflection-print duplication, now fixed the same way.ApplyHeatlookup losing its only grounding once the dump was removed) is fixed with a one-sentence bridge pointing to Chapter 4's own existing action-flow diagram, rather than reopening the removed excerpt.Process
Two-phase: a 9-agent read-only survey (
decisions/diagram-text-integration-survey.md) produced the literal before/after content for every cell touched; a 9-task implementation plan (docs/superpowers/plans/2026-10-02-diagram-text-integration-phase-b-plan.md) then executed each chapter as its own contract — independent worktree, builder on Sonnet 5, reviewer on Opus 5.5, merged only after an independent PASS. Every chapter's notebooks were executed end-to-end (jupyter nbconvert --execute) as part of that review, and the five dump-trimmed chapters each got an additionalsimulated-learnerreader-comprehension spot-check.Test plan
uv run pytest tests/ glossary/tests/ -q— 444 passed, 7 deselected (verified after every merge)uv run python scripts/check_construction.py --check— clean after every mergemodels/chNN-cumulative.sysmlsimulated-learnerspot-check on all 5 whole-dump-trim chapters (Ch2, Ch3, Ch4, Ch5, Ch6) — PASSnpx mystmd start --execute) — 59 pages built, server serves 200s