superlab 0.1.65 → 0.1.67

package/package-assets/shared/lab/.managed/templates/write-iteration.md

@@ -67,33 +67,6 @@
 - Any discouraged move kept and why:
 - Any banned move found:
 
-## Review Issue Bundle
-
-- Issue bundle path:
-- New issues:
-- Resolved issues:
-- Open issues:
-- Quote-backed findings recorded:
-- Script-backed findings separated from judgment-backed findings:
-
-## Re-Audit Status
-
-- Previous issue bundle compared:
-- Fully addressed root causes:
-- Partially addressed root causes:
-- Not addressed root causes:
-- New root causes:
-- Which root-cause issues block further prose polish:
-
-## Reference Template Intake
-
-- Reference sources used:
-- Aggregate template playbook:
-- Section templates consulted:
-- Visual/table templates consulted:
-- Multi-template reproduction plan:
-- Structure-only reuse boundary:
-
 ## Table Semantics
 
 - Metrics promised in Method:
@@ -142,6 +115,21 @@
 - Was workflow-language paper layer included in the exported/pushed bundle:
 - If workflow-language was omitted, why was canonical-only export acceptable:
 
+## Reference Structure Consumption
+
+- Was reference-guided deep writing triggered:
+- Reference sources used:
+- Reference consumption plan path:
+- Were section/subsection slots mapped before prose:
+- Were paragraph roles mapped before prose:
+- Were table/figure roles mapped before prose:
+- Which reference slots were adopted:
+- Which reference slots were waived and why:
+- Did the round avoid copying reference wording, claims, metrics, captions, or conclusions:
+- Did final prose avoid service-style or AI-assistant meta language:
+- Did final prose avoid workflow-only placeholder language:
+- Validator command and result:
+
 ## Decision
 
 - Continue or stop:

package/package-assets/shared/skills/lab/SKILL.md

@@ -246,6 +246,9 @@ Use this skill when the user invokes `/lab:*` or asks for the structured researc
 - Keep one canonical natural-language paper-facing name per concept.
 - Once a paper-facing model or ablation label is chosen, reuse the canonical label instead of replacing it with a narrative alias in later prose, tables, or captions.
 - Before drafting or polishing, check the current section block in `skills/lab/references/paper-writing/section-style-policies.md` and follow its encouraged, discouraged, and banned expression lists.
+- When the user provides reference PDFs, paper URLs, local reference-paper paths, or asks to write by reference, stay within `/lab:write` but switch to reference-guided deep writing: extract structure, map section/subsection slots, paragraph roles, and table/figure roles to the current paper, record the mapping, and only then draft prose.
+- Reference-guided writing may reuse structure, paragraph roles, asset placement, and bridge logic, but must not copy reference wording, claims, metrics, captions, or conclusions.
+- Keep service-style, AI-assistant meta language, and workflow-only placeholder language out of manuscript prose, captions, table notes, and paper-facing analysis assets.
 - Before any additional tighten, compress, or polish pass on the same section, run a section-level acceptance gate first.
 - The section-level acceptance gate must explicitly check canonical naming consistency, adjacent-section consistency, claim/metric/ranking consistency with evidence, local clarity, local concision, and section-style compliance.
 - If the current section still contains a banned expression or banned rhetorical move from the section-style policy, the round has not passed the section-level acceptance gate.

package/package-assets/shared/skills/lab/references/paper-writing/examples/experiments/figure-placeholder-and-discussion.md

@@ -10,9 +10,11 @@ attachment.
 \begin{figure}[t]
 \centering
 \fbox{\rule{0pt}{1.55in}\rule{0.92\linewidth}{0pt}}
-\caption{Method overview. Figure intent: show the full pipeline, highlight the
-boundary between the structured scoring module and the post-hoc calibration
-stage, and make the train-time versus inference-time data flow easy to inspect.}
+% Authoring note: the final visual should show the full pipeline, highlight the
+% module boundary, and make train-time versus inference-time data flow visible.
+\caption{Method overview of the proposed model. The diagram separates the
+structured scoring module from the post-hoc calibration stage and indicates the
+data flow used during training and inference.}
 \label{fig:method-overview}
 \end{figure}
 ```
@@ -23,9 +25,11 @@ stage, and make the train-time versus inference-time data flow easy to inspect.}
 \begin{figure}[t]
 \centering
 \fbox{\rule{0pt}{1.55in}\rule{0.92\linewidth}{0pt}}
-\caption{Benchmark-level results overview. Figure intent: summarize the trend
-across datasets, show error bars or confidence intervals, and reveal whether the
-main gain is stable or dominated by one benchmark.}
+% Authoring note: the final visual should summarize cross-dataset trends and
+% uncertainty without introducing a claim absent from the result tables.
+\caption{Benchmark-level results overview. The plot summarizes the primary
+metric across datasets and shows whether the reported gain is stable across the
+evaluated settings.}
 \label{fig:results-overview}
 \end{figure}
 ```

package/package-assets/shared/skills/lab/references/paper-writing/examples/experiments-examples.md

@@ -8,7 +8,7 @@ glue, not just checklists.
 
 1. Main results should live in a real `table` environment.
 2. Ablations should live in a separate `table` environment.
-3. Method and experiments should each have at least one figure placeholder with an explicit `Figure intent`.
+3. Method and experiments should each have at least one figure placeholder with a reader-ready caption and any authoring notes kept outside the caption.
 4. Captions should explain the table or figure message briefly; longer interpretation belongs in prose.
 
 ## Example Files

package/package-assets/shared/skills/lab/references/paper-writing/section-style-policies.md

@@ -30,6 +30,8 @@ These are paper-facing defaults. They are not project-specific branding rules.
 - Roadmap prose such as "In this paper, we first..., then..., finally...".
 - Reviewer-facing instructions such as "the reader can see" or "as shown clearly below".
 - Unbounded superiority claims such as "universally", "always", or "in every setting".
+- Service-style or AI-assistant meta language such as "用户说", "按你的要求", "我来解释", "let me explain", or "as requested by the user".
+- Workflow-only placeholder language such as "图的意图", "资产意图", "占位符", "workflow-language", or "sync this wording".
 
 ## Introduction
 
@@ -51,6 +53,8 @@ These are paper-facing defaults. They are not project-specific branding rules.
 - Empty macro-importance claims such as "this problem is increasingly critical" with no concrete consequence.
 - Marketing-style first-claim language such as "revolutionary", "game-changing", or "unprecedented" without evidence.
 - Paragraphs that only praise the paper instead of stating the research gap.
+- Service-style or AI-assistant meta language such as "用户说", "按你的要求", "我来解释", "let me explain", or "as requested by the user".
+- Workflow-only placeholder language such as "图的意图", "资产意图", "占位符", "workflow-language", or "sync this wording".
 
 ## Related Work
 
@@ -71,6 +75,8 @@ These are paper-facing defaults. They are not project-specific branding rules.
 - Laundry-list paragraphs that only say "X does..., Y does..., Z does..." with no comparison.
 - Claims that related work is weak or obsolete without specifying the missing capability.
 - Hiding the closest prior work behind broad category language.
+- Service-style or AI-assistant meta language such as "用户说", "按你的要求", "我来解释", "let me explain", or "as requested by the user".
+- Workflow-only placeholder language such as "图的意图", "资产意图", "占位符", "workflow-language", or "sync this wording".
 
 ## Method
 
@@ -92,6 +98,8 @@ These are paper-facing defaults. They are not project-specific branding rules.
 - Marketing-style or self-promotional wording such as "elegant", "powerful", "dramatically stronger", or "significantly outperforms prior methods" when used as prose decoration rather than evidence-backed result reporting.
 - Explaining the method by saying it is "better", "stronger", or "more advanced" without saying how it works.
 - Introducing new narrative aliases for canonical model or ablation labels after they have already been locked.
+- Service-style or AI-assistant meta language such as "用户说", "按你的要求", "我来解释", "let me explain", or "as requested by the user".
+- Workflow-only placeholder language such as "图的意图", "资产意图", "占位符", "workflow-language", or "sync this wording".
 
 ## Experiments
 
@@ -114,6 +122,8 @@ These are paper-facing defaults. They are not project-specific branding rules.
 - Self-evaluations such as "结果也很清楚", "the defense results are very clear", or "the table is self-explanatory".
 - Layout-process commentary in scientific prose, such as "由于表列较多，这里采用页宽自适应排版" or "we use page-width adaptive layout here".
 - Claims that a table "proves" something when the evidence only supports a bounded empirical result.
+- Service-style or AI-assistant meta language such as "用户说", "按你的要求", "我来解释", "let me explain", or "as requested by the user".
+- Workflow-only placeholder language such as "图的意图", "资产意图", "占位符", "workflow-language", or "sync this wording".
 
 ## Conclusion
 
@@ -134,3 +144,5 @@ These are paper-facing defaults. They are not project-specific branding rules.
 - Introducing new evidence, new experiments, or new mechanism claims.
 - Expanding the paper's scope beyond what the experiments support.
 - Ending with generic hype such as "this opens a new era" or "this will broadly transform the field".
+- Service-style or AI-assistant meta language such as "用户说", "按你的要求", "我来解释", "let me explain", or "as requested by the user".
+- Workflow-only placeholder language such as "图的意图", "资产意图", "占位符", "workflow-language", or "sync this wording".

package/package-assets/shared/skills/lab/stages/write.md

@@ -69,6 +69,26 @@ Run these on every round:
 - reviewer pass -> `skills/lab/references/paper-writing/paper-review.md`
 - section-specific style policy -> `skills/lab/references/paper-writing/section-style-policies.md` (load the block matching the current section)
 
+## Reference-Guided Deep Write
+
+Trigger this mode automatically when the user provides reference PDFs, paper URLs, local reference-paper paths, a template paper, or asks to "参考" papers while continuing `/lab:write`.
+
+This mode is still `/lab:write`; do not introduce a separate user command. It increases the depth of the write round instead of asking the user to remember another stage.
+
+Required sequence:
+
+1. Extract or refresh reference-paper structure with `.lab/.managed/scripts/extract_reference_paper_structure.py` when new reference PDFs or URLs are provided.
+2. Read the generated `section-map.json`, `section-logic.md`, `paragraph-roles.json`, `visual-assets.json`, and aggregate `section-templates/<section>.json` for the current section.
+3. Write `.lab/writing/reference-patterns/consumption-plan/<section>.md` from `.lab/.managed/templates/reference-consumption-plan.md` before drafting prose.
+4. Map reference section/subsection slots, paragraph roles, table roles, figure roles, and bridge logic to the current paper's evidence and active paper layer.
+5. Explicitly waive any reference slot that is not used, with a reason tied to current evidence or scope.
+6. Reuse structure only. Do not copy wording, claims, metrics, captions, or conclusions from the reference papers.
+7. Run `.lab/.managed/scripts/validate_reference_consumption.py --section <section> --section-file <section-file> --mode draft` after drafting; final-draft or export rounds must use `--mode final`.
+
+For experiments, the consumption plan must cover or explicitly waive dataset description/statistics, split protocol, baseline setup, metric definition, implementation details, main results, ablation, and sensitivity/robustness. Dataset description, split protocol, baseline setup, metric definition, and main results are core slots and should be mapped rather than waived in normal empirical papers.
+
+Do not enter prose polish until the current section has passed the reference-consumption check or has a recorded reason why reference-guided writing was not triggered.
+
 ## Small-Step Writing Rules
 
 - Change one section or one clearly bounded subsection per round.
@@ -113,14 +133,8 @@ Run these on every round:
 - Load only the current section guide. Do not load every section guide at once.
 - Reuse example-bank structure, paragraph roles, sentence logic, and paper-facing LaTeX asset patterns when examples are bundled, but never copy wording verbatim.
 - Treat example cites and example file names as writing references, not as evidence for the current paper.
-- When the user provides local PDFs, PDF URLs, HTML pages, or reference papers while invoking `/lab:write`, run `.lab/.managed/scripts/extract_reference_paper_structure.py --output-dir .lab/writing/reference-patterns <sources...>` before drafting unless an up-to-date `.lab/writing/reference-patterns/aggregate-template-playbook.md` already covers those exact sources.
-- Treat reference-paper intake as an internal write capability, not a separate user command. The user should still only need `/lab:write`; do not ask them to learn another workflow.
-- The purpose of reference-paper intake is to help `/lab:write` reproduce mature multi-template writing structure: section slots, paragraph roles, argument sequence, table and figure functions, placement logic, and bridge sentences.
-- Use at least two compatible reference templates when available. If only one reference is available, mark it as a single-template pattern and avoid treating it as a universal standard.
-- For every reference table or figure, extract what reader question it answers, which section/subsection it supports, why it is placed there, what the prose before it should do, and what the prose after it should explain.
-- When drafting from reference templates, reproduce structure and logic only. Do not copy wording, claims, metrics, baselines, data, captions, or conclusions from reference papers.
-- Before drafting a section from reference templates, read `.lab/writing/reference-patterns/aggregate-template-playbook.md`, the matching file under `.lab/writing/reference-patterns/section-templates/`, and the matching visual/table template under `.lab/writing/reference-patterns/visual-templates/` when the section uses tables or figures.
 - Build a compact mini-outline before prose.
+- When reference-guided deep-write is triggered, build the reference consumption plan before the mini-outline so the outline is based on mapped section slots rather than generic prose flow.
 - Academic readability standards are the same in `workflow_language` and `paper_language`; changing languages must not lower external-reader clarity.
 - If the current round introduces or revises key terms, abbreviations, metric names, mechanism names, or system labels, explain them at first mention by briefly stating what they are and why they matter here.
 - First mention should use the full form. If a short form or acronym will be reused later, define it at first mention as `Full Form (Short Form)` before switching to the short form.
@@ -133,10 +147,6 @@ Run these on every round:
 - Before any additional tighten, compress, or polish pass on the same section, run a section-level acceptance gate first.
 - The section-level acceptance gate is passed only when canonical naming consistency, adjacent-section consistency, claim, metric, and ranking consistency with the current evidence, local clarity, local concision, and section-style compliance are all explicitly checked and no unresolved blocker remains.
 - If the current section still contains a banned expression or banned rhetorical move from `section-style-policies.md`, the round has not passed the section-level acceptance gate.
-- If reviewer notes, validator warnings, or prior write rounds produced issues, record them as a review issue bundle in the write-iteration artifact before further polishing.
-- Review issue bundles should separate script-backed findings from judgment-backed findings, preserve the source quote or local pointer when available, and track whether each issue is new, resolved, open, or superseded.
-- Before continuing prose polish after a review issue bundle exists, run a re-audit pass that compares the current draft against previous root causes and records fully addressed, partially addressed, not addressed, and newly introduced root causes.
-- Do not answer a review issue by merely changing wording around it. Fix the underlying section structure, evidence support, terminology definition, or asset/table linkage that caused the issue.
 - If the current round changes the paper's canonical experiment or evaluation protocol (for example split ratio, train/test size, seed or split count, benchmark set, or main-table evaluation contract), treat it as a canonical protocol replacement unless the user explicitly scopes it as supplementary or appendix-only.
 - A canonical protocol replacement requires a paper-wide impact audit before more polishing: identify stale sections and assets across Abstract, Introduction, Method, Experiments, Conclusion, tables, figures, analysis assets, and `.lab/writing/plan.md`, then update the plan and highest-impact stale targets first.
 - When a paper-wide impact audit is still open, default the next write action to the highest-impact canonical stale section or asset instead of polishing the same section again.
@@ -150,6 +160,8 @@ Run these on every round:
 - Do not use labels containing `_` or `-` in reader-facing prose.
 - Keep internal identifiers, config keys, and experiment package labels out of reader-facing prose unless they are mapped once for the reader and then moved back out of prose.
 - Do not use unexplained terminology density as a substitute for academic tone.
+- Keep service-style or AI-assistant meta language out of manuscript prose. Phrases such as "用户说", "按你的要求", "我来解释", "下面我", "this version", or "as requested by the user" belong in workflow notes, not in paper-facing sections, captions, table notes, or analysis assets.
+- Keep workflow-only placeholder language out of manuscript prose. Phrases such as "图的意图", "资产意图", "占位符", "workflow-language", "translation layer", or "sync this wording" belong in authoring artifacts, not in reader-facing LaTeX.
 - Main tables must be locally self-contained: a reader looking at the table title, header, note, and adjacent introduction or interpretation should be able to determine what each row represents, what each column measures, the metric direction, and any relevant unit, denominator, or event condition.
 - Short headers remain allowed, but they must be resolved locally through the same table's caption or table note instead of forcing the reader to chase the Method section.
 - If the Method or Experiments prose says the paper reports a metric family, the main table set must either expose those metrics directly or explicitly mark the missing ones as appendix-only and explain why.
@@ -172,9 +184,9 @@ Run these on every round:
 - When the repository workflow config is available, the paper-plan validator also checks that `.lab/writing/plan.md` stays in `workflow_language` instead of silently drifting into another language.
 - If the paper-plan validator fails, stop and fill `.lab/writing/plan.md` first instead of drafting prose.
 - During ordinary draft rounds, run `.lab/.managed/scripts/validate_section_draft.py --section <section> --section-file <section-file> --mode draft` and `.lab/.managed/scripts/validate_paper_claims.py --section-file <section-file> --mode draft` after revising the active section.
+- If reference-guided deep-write was triggered, also run `.lab/.managed/scripts/validate_reference_consumption.py --section <section> --section-file <section-file> --mode draft` after revising the active section.
 - Treat draft-round output from the section and claim validators as warnings that must be recorded and addressed in the write-iteration artifact, not as immediate stop conditions.
 - If the active section already lives under a paper-layer `sections/` directory, the draft section validator should also warn when the neighboring required figure or analysis placeholder files are still missing from that same paper layer.
-- For experiment sections, treat prose-only performance claims, unnamed generic comparator phrases, repeated split/seed protocols without variance disposition, and result paragraphs without concrete metric/table anchors as section warnings that must be fixed before more prose-only polishing.
 - For each subsection, explicitly include motivation, design, and technical advantage when applicable.
 - Avoid a writing style that reads like incremental patching of a naive baseline.
 - Keep terminology stable across the full paper.
@@ -195,14 +207,12 @@ Run these on every round:
   - `<deliverables_root>/paper/analysis/analysis-asset.tex`
 - Table assets must use paper-facing LaTeX structure with `booktabs`, caption, label, and consistent precision.
 - Table assets must also include a local table note that explains row meaning, column meaning, metric definitions, comparison scope, and any important caveat.
-- Table assets must avoid vertical rules, `\hline`, and `\cline`; use `booktabs` rules and whitespace instead.
-- Table captions should appear before the tabular body so the table can be read top-down in manuscript order.
-- Numeric precision should be consistent within each metric column unless the table note explains a deliberate exception.
 - Table assets must not rely on aggressive width hacks by default; if width control is still needed after table redesign, document it locally and keep it readable.
-- Figure placeholders must explain what the final figure should show and why the reader needs it.
+- Figure placeholders may record what the final figure should show and why the reader needs it in authoring comments, the paper plan, or the write-iteration artifact, but the caption itself must remain paper-facing and must not contain "Figure intent", "图的意图", "asset intent", "占位符", or similar workflow language.
 - Core asset coverage for a paper-facing final draft should include a problem-setting or teaser figure, a method overview figure, a results overview figure, a main-results table, an ablation table, and one additional analysis asset.
 - Keep `.lab/writing/plan.md` synchronized with the current table plan, figure plan, citation plan, and section-to-asset map whenever manuscript assets change.
 - For final-draft or export rounds, run `.lab/.managed/scripts/validate_section_draft.py --section <section> --section-file <section-file> --mode final` and `.lab/.managed/scripts/validate_paper_claims.py --section-file <section-file> --mode final` before accepting the round.
+- If reference-guided deep-write was triggered, run `.lab/.managed/scripts/validate_reference_consumption.py --section <section> --section-file <section-file> --mode final` before accepting the final-draft or export round.
 - If the final-round section or claim validators fail, keep editing the affected section until it passes; do not stop at asset-complete but rhetorically weak or unsafe prose.
 - Final-round section validation should fail when a section in the paper layer references required figure or analysis placeholders but the neighboring asset files are still missing from that layer.
 - Run `.lab/.managed/scripts/validate_manuscript_delivery.py --paper-dir <deliverables_root>/paper` before accepting a final-draft or export round.
@@ -215,8 +225,6 @@ Run these on every round:
 - When a round introduces or revises key terms, include a compact terminology note in the user-facing round summary and record the terminology-clarity self-check in the write-iteration artifact.
 - Record the section-level acceptance gate in the write-iteration artifact before recommending further tightening on the same section.
 - Record section-style policy compliance, any retained discouraged move, and any banned move found in the write-iteration artifact.
-- Record the review issue bundle and re-audit status in the write-iteration artifact whenever the round follows reviewer notes, validator warnings, or prior failed writing rounds.
-- Record the reference template intake in the write-iteration artifact whenever the round uses PDFs, URLs, or `.lab/writing/reference-patterns/` artifacts: sources used, aggregate playbook path, section templates consulted, visual/table templates consulted, multi-template reproduction plan, and structure-only reuse boundary.
 - Record the round target layer in the write-iteration artifact as `canonical manuscript`, `workflow-language paper layer`, or `both`.
 - If workflow-language was active and the round still targeted the canonical manuscript, record why canonical-only writing was acceptable in the write-iteration artifact.
 - If both layers were edited, record why the cross-language sync was required and whether it was explicitly requested by the user or required by final-draft/export finalization.
@@ -237,6 +245,7 @@ Run these on every round:
 - `.lab/writing/framing.md`
 - `.lab/writing/plan.md`
 - `.lab/writing/terminology-glossary.md`
+- `.lab/writing/reference-patterns/consumption-plan/<section>.md` when reference-guided deep-write is triggered
 - `.lab/writing/iterations/<n>.md`
 - `<deliverables_root>/paper/main.tex`
 - `<deliverables_root>/paper/references.bib`

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "superlab",
-  "version": "0.1.65",
+  "version": "0.1.67",
   "description": "Strict /lab research workflow installer for Codex and Claude",
   "keywords": [
     "codex",