scientify 3.0.0 → 3.2.0

package/skills/write-paper/references/style-banlist.md

@@ -0,0 +1,81 @@
+# Style Banlist
+
+This is not a universal forbidden-word list.
+
+Use it as a risk map: the higher the claim pressure, the more these words need explicit support nearby.
+
+## High-Risk Praise Words
+
+These often create fake confidence when they are not tied to evidence in the same sentence or the next sentence:
+
+- `significant`
+- `substantial`
+- `strong`
+- `robust`
+- `effective`
+- `promising`
+- `remarkable`
+- `novel`
+- `comprehensive`
+- `state-of-the-art`
+
+## High-Risk Hedge Words
+
+These are useful for interpretation, but risky when they blur what is actually observed:
+
+- `may`
+- `can`
+- `could`
+- `potentially`
+- `suggests`
+- `appears to`
+- `likely`
+- `in general`
+- `overall`
+
+## Vague Result Phrases
+
+These usually need to be rewritten into metric- or evidence-based language:
+
+- `shows advantages`
+- `performs well`
+- `works effectively`
+- `achieves competitive results`
+- `delivers better performance`
+- `improves overall quality`
+- `demonstrates robustness`
+- `has broad applicability`
+
+## Rewrite Moves
+
+When a sentence sounds too soft, too grand, or too vague, prefer one of these moves:
+
+- add a metric
+- add a baseline
+- add a protocol or guardrail
+- add a source anchor such as `claim_id`, figure, or table
+- add a scope or evidence boundary
+- split observation from interpretation
+
+Examples:
+
+- `strong improvement` -> `17.53% mean TTFT gain vs INT4-FIFO`
+- `robust behavior` -> `maintains quality_penalty_mean <= 0.02 under the stated protocol`
+- `promising result` -> `improves bytes/request under simulator evaluation, but lacks runtime validation`
+
+## Usage Rules
+
+- If one of the high-risk praise words appears in a result sentence, the same sentence or the next sentence should also contain a metric, baseline, or explicit evidence anchor.
+- Hedge words are acceptable when they clearly mark interpretation, uncertainty, or future-facing reasoning.
+- Do not use hedge words to hide whether a result is actually observed or merely inferred.
+- If a sentence can be made more precise by adding a number, comparison target, or boundary, do that instead of adding emphasis.
+
+## Surface-Aware Notes
+
+Different writing surfaces tolerate different kinds of language:
+
+- high-pressure surfaces such as abstracts, headline result sentences, captions, and opening-page summaries need the strictest wording
+- medium-pressure surfaces such as result discussion and comparison paragraphs can carry some interpretation, but still need explicit evidence anchors
+- lower-pressure surfaces such as discussion, scope notes, and future-work paragraphs can use more hedging, as long as they do not rewrite observed results into broader claims
+
+The key question is not “what section am I in?” but “how much claim pressure does this sentence carry?”

package/skills/write-review-paper/SKILL.md

@@ -60,6 +60,8 @@ Create `review/reading_plan.md`:
 - [ ] ...
 ```
 
+Every entry must keep `paper_id` and the selection reason. Do not write a reading plan that only lists titles.
+
 ### 1.2 Reading Notes Template
 
 For each paper, create `review/notes/{paper_id}.md` using template in `references/note-template.md`.
@@ -75,10 +77,10 @@ Create `review/comparison.md`:
 ```markdown
 # Method Comparison
 
-| Paper | Year | Category | Key Innovation | Dataset | Metric | Result |
-|-------|------|----------|----------------|---------|--------|--------|
-| [A]   | 2023 | Data-driven | ... | ... | RMSE | 0.05 |
-| [B]   | 2022 | Hybrid | ... | ... | RMSE | 0.08 |
+| Paper | Year | Category | Key Innovation | Dataset | Metric | Result | Evidence / Source |
+|-------|------|----------|----------------|---------|--------|--------|-------------------|
+| [A]   | 2023 | Data-driven | ... | ... | RMSE | 0.05 | `review/notes/{paper_id}.md` |
+| [B]   | 2022 | Hybrid | ... | ... | RMSE | 0.08 | `review/notes/{paper_id}.md` |
 ```
 
 ### 2.2 Timeline Analysis
@@ -144,6 +146,8 @@ Create `review/draft.md` using template in `references/survey-template.md`.
 
 Key sections: Abstract → Introduction → Background → Taxonomy → Comparison → Datasets → Future Directions → Conclusion
 
+At the end of each major section, add one short summary sentence that clearly reflects the evidence already written in `review/notes/` or `review/comparison.md`.
+
 ### 3.2 Thesis Literature Review Template
 
 For a thesis chapter:
@@ -192,6 +196,8 @@ For a thesis chapter:
 3. **时态混乱** - 描述方法用现在时，描述实验结果用过去时
 4. **过度引用** - 不是每句话都需要引用
 5. **遗漏重要工作** - 确保覆盖领域的奠基性工作
+6. **Body text detached from notes** - Do not write conclusions into the draft unless they already appear in notes / comparison
+7. **Trend written as certainty** - When evidence is not stable, frame it as an observation or discussion rather than a firm conclusion
 
 ---