research-copilot 0.1.0

package/lib/skills/builtin/research-grants/references/specific_aims_guide.md

@@ -0,0 +1,458 @@
+# NIH Specific Aims Page: The Complete Guide
+
+## Overview
+
+The **Specific Aims page** is the most important page of your entire NIH grant application. It's the first thing reviewers read, often determines their initial impression, and may be the only page read by some panel members before scoring begins.
+
+**Length**: Exactly 1 page
+**Margins**: 0.5 inches (all sides)
+**Font**: 11-point Arial, Helvetica, or similar (no smaller)
+**Line spacing**: Must be readable
+
+**Purpose**: 
+- Communicate your research vision clearly and compellingly
+- Establish significance and innovation
+- Demonstrate feasibility
+- Show that you can accomplish meaningful work in the proposed timeframe
+- Make reviewers excited to fund your work
+
+## Anatomy of a Specific Aims Page
+
+### Essential Components (in order)
+
+1. **Opening Hook** (2-4 sentences)
+2. **Gap/Problem Statement** (2-4 sentences)  
+3. **Long-Term Goal** (1 sentence)
+4. **Objective** (1-2 sentences)
+5. **Central Hypothesis** (1 sentence) [or Research Questions]
+6. **Rationale** (2-3 sentences with preliminary data mention)
+7. **Specific Aims** (2-4 aims, ~½ page total)
+8. **Expected Outcomes and Impact** (2-4 sentences)
+
+## Detailed Structure
+
+### Opening Paragraph: The Hook
+
+**Purpose**: Establish importance and grab attention
+
+**What to include**:
+- Broad context (disease burden, biological importance, technological need)
+- Epidemiological data or statistics that establish scale
+- Why this problem matters for health or science
+- Create urgency
+
+**Length**: 2-4 sentences
+
+**Writing tips**:
+- Start strong with compelling statement
+- Use concrete numbers (prevalence, mortality, costs)
+- Avoid jargon in first sentence
+- Make it accessible to non-specialists on panel
+
+**Examples**:
+
+*Clinical Example*:
+"Pancreatic ductal adenocarcinoma (PDAC) is the third leading cause of cancer death in the United States, with a devastating 5-year survival rate of only 11%. Despite decades of research, therapeutic options remain limited, and most patients present with advanced, unresectable disease. The lack of effective early detection methods and targeted therapies represents a critical unmet medical need affecting over 62,000 Americans diagnosed annually."
+
+*Basic Science Example*:
+"Mitochondrial dysfunction is a hallmark of aging and age-related diseases, yet the mechanisms linking mitochondrial decline to cellular senescence remain poorly understood. Emerging evidence suggests that mitochondrial-nuclear communication pathways play a central role in longevity determination across species, from yeast to mammals. Understanding how cells sense and respond to mitochondrial stress could reveal new therapeutic targets for age-related diseases affecting millions worldwide."
+
+### Second Paragraph: Gap and Context
+
+**Purpose**: Define what's known, what's unknown, and why it matters
+
+**What to include**:
+- Current state of knowledge (brief literature context)
+- Specific gap or barrier to progress
+- Why this gap is critical to address
+- Why current approaches are insufficient
+
+**Length**: 3-5 sentences
+
+**Structure**:
+1. What we know (1-2 sentences)
+2. What we don't know / what's limiting progress (1-2 sentences)
+3. Why this gap matters (1 sentence)
+
+**Examples**:
+
+"Prior studies have identified numerous genetic mutations associated with PDAC development, including KRAS, TP53, SMAD4, and CDKN2A. However, the tumor microenvironment (TME), comprising immune cells, fibroblasts, and extracellular matrix, is increasingly recognized as a critical determinant of therapeutic resistance. Current models fail to recapitulate the complex TME architecture and cell-cell interactions that drive therapy resistance in vivo, limiting our ability to develop effective treatments. Understanding how the TME protects tumor cells from chemotherapy is essential for designing combination therapies that overcome resistance."
+
+### Third Paragraph: Long-Term Goal, Objective, Hypothesis, Rationale
+
+**Purpose**: Set up your specific approach and justification
+
+**Structure**:
+
+**Long-Term Goal** (1 sentence):
+- Your overarching research program direction
+- Broader than this specific proposal
+- Provides context for this work
+
+*Example*: "The long-term goal of our research is to elucidate the molecular mechanisms by which the tumor microenvironment promotes therapeutic resistance in pancreatic cancer."
+
+**Objective** (1-2 sentences):
+- Specific objective of THIS grant
+- What you will accomplish in 3-5 years
+- More focused than long-term goal
+
+*Example*: "The objective of this application is to define the role of cancer-associated fibroblasts (CAFs) in mediating gemcitabine resistance and to develop combination therapies targeting CAF-tumor interactions."
+
+**Central Hypothesis** (1 sentence):
+- Testable prediction
+- Should unify the specific aims
+- Based on preliminary data or logical reasoning
+- Clear and specific
+
+*Example*: "Our central hypothesis is that CAF-secreted factors activate protective autophagy in tumor cells, conferring resistance to gemcitabine, and that dual inhibition of CAF signaling and autophagy will restore drug sensitivity."
+
+**Alternative: Research Questions** (if hypothesis-testing isn't appropriate):
+- 2-3 focused questions
+- Should correspond to specific aims
+
+*Example*: "This project will address the following questions: (1) What factors secreted by CAFs promote tumor cell survival during chemotherapy? (2) How do tumor cells integrate CAF signals to activate protective responses? (3) Can targeting CAF-tumor interactions enhance therapeutic efficacy in preclinical models?"
+
+**Rationale** (2-3 sentences):
+- Why you think the hypothesis is true
+- Mention key preliminary data (very briefly)
+- Logical basis for your approach
+- Why this approach will work
+
+*Example*: "This hypothesis is based on our preliminary data showing that CAF-conditioned medium protects tumor cells from gemcitabine-induced apoptosis by 60% (Fig. 1), and that this protection is blocked by autophagy inhibitors (Fig. 2). Proteomic analysis of CAF secretomes identified 15 candidate factors enriched in drug-resistant contexts (Table 1). These findings suggest a targetable pathway linking CAF signaling to tumor cell survival that could be exploited therapeutically."
+
+### Specific Aims (Main Section)
+
+**How many aims**: 2-4 aims (3 is most common for R01)
+- **Too few (1)**: Insufficient work, appears risky
+- **Just right (2-3)**: Focused, achievable, synergistic
+- **Too many (4+)**: Overly ambitious, unlikely to complete
+
+**Structure for each aim**:
+1. **Aim Statement** (1-2 sentences, bold or underlined)
+2. **Rationale and Background** (1-3 sentences)
+3. **Working Hypothesis** (1 sentence, if applicable)
+4. **Approach Summary** (2-4 sentences)
+5. **Expected Outcomes and Interpretation** (1-2 sentences)
+
+**Length per aim**: ~4-6 sentences (¼ to ⅓ page)
+
+**Relationships between aims**:
+- **Independent**: Failure of one aim doesn't doom the others
+- **Synergistic**: Aims build on each other or address complementary questions
+- **Progressive**: Aim 1 enables Aim 2, Aim 2 enables Aim 3 (be careful—creates risk)
+
+#### Example Aim Structure:
+
+**Aim 1: Identify CAF-secreted factors that mediate gemcitabine resistance.**
+
+*Rationale*: CAF-conditioned medium confers significant protection against gemcitabine (Fig. 1), suggesting secreted factors are responsible. We have identified 15 candidate proteins enriched in CAF secretomes from resistant versus sensitive contexts (Table 1). 
+
+*Working Hypothesis*: CAFs secrete specific growth factors and cytokines (including IL-6, CXCL12, and HGF) that activate pro-survival pathways in tumor cells.
+
+*Approach*: We will (1) validate candidate factors using neutralizing antibodies in co-culture assays, (2) measure activation of downstream signaling pathways (STAT3, PI3K/AKT, MAPK) in tumor cells, and (3) perform CRISPR screens in CAFs to identify factors required for resistance phenotype. We will use patient-derived CAFs and tumor cells to ensure clinical relevance.
+
+*Expected Outcomes*: We expect to identify 3-5 CAF-secreted factors sufficient and necessary for gemcitabine resistance, and define their signaling mechanisms. These will serve as therapeutic targets for Aims 2-3.
+
+---
+
+**Aim 2: Determine the mechanisms by which CAF signals activate protective autophagy in tumor cells.**
+
+*Rationale*: Our data show that CAF-mediated resistance requires autophagy (Fig. 2), but the signaling pathways linking CAF factors to autophagy activation remain unknown.
+
+*Working Hypothesis*: CAF-secreted factors activate mTOR-independent autophagy through AMPK and ULK1 phosphorylation.
+
+*Approach*: We will (1) measure autophagy flux in tumor cells exposed to CAF factors using LC3 turnover assays and electron microscopy, (2) define signaling pathways using phosphoproteomic analysis and pharmacologic inhibitors, and (3) validate pathways using genetic knockdowns (shRNA/CRISPR) of key nodes. Studies will be performed in 2D and 3D co-culture systems.
+
+*Expected Outcomes*: We will define the signaling cascade from CAF factors to autophagy activation, identifying druggable nodes for combination therapy. Results will inform Aim 3 therapeutic strategies.
+
+---
+
+**Aim 3: Evaluate combination therapies targeting CAF-tumor interactions in preclinical models.**
+
+*Rationale*: Single-agent therapies targeting CAFs or autophagy have shown limited efficacy clinically, suggesting combination approaches are needed.
+
+*Working Hypothesis*: Dual inhibition of CAF signaling and autophagy will synergistically restore gemcitabine sensitivity in vivo.
+
+*Approach*: Using patient-derived xenograft (PDX) models and genetically engineered mouse models (GEMM) of PDAC, we will test combinations of (1) gemcitabine + CAF pathway inhibitors identified in Aim 1, (2) gemcitabine + autophagy inhibitors, and (3) triple combinations. We will assess tumor growth, survival, and mechanism (IHC, RNA-seq) in n=10-15 mice per group.
+
+*Expected Outcomes*: We expect combination therapies will reduce tumor growth by ≥60% compared to gemcitabine alone, with synergistic effects. The most effective regimen will be advanced toward clinical translation through an investigator-initiated trial (we have IND-enabling resources available at our institution).
+
+### Closing Paragraph: Impact and Significance
+
+**Purpose**: Leave reviewers with enthusiasm and clear understanding of importance
+
+**What to include**:
+- Expected outcomes of the overall project
+- How findings will advance the field
+- Positive impact on health or science
+- Next steps or future directions
+- Why this matters
+
+**Length**: 2-4 sentences
+
+**Writing tips**:
+- Be confident but not arrogant
+- Connect back to opening (full circle)
+- Emphasize transformative potential
+- Avoid over-promising
+
+**Examples**:
+
+"The proposed research is significant because it will define a novel mechanism of chemotherapy resistance in pancreatic cancer and identify new therapeutic targets to overcome this resistance. Results will provide mechanistic insights into CAF-tumor interactions that drive drug resistance, immediately applicable to clinical trial design. We expect findings will enable rational design of combination therapies that improve outcomes for PDAC patients, who currently have few effective treatment options. This work will establish new paradigms for targeting the tumor microenvironment in solid cancers."
+
+## Writing Principles
+
+### Clarity and Accessibility
+
+**Write for a mixed audience**:
+- Some panel members will be experts in your area
+- Others will be in related but not identical fields
+- Program officers and council members will read it
+- Some reviewers will only read this page before scoring
+
+**Strategies**:
+- Define technical terms at first use
+- Explain abbreviations (except very common ones)
+- Use clear, direct language
+- Avoid excessive jargon
+- Make logical flow obvious
+
+### Confidence Without Arrogance
+
+**Confident** ✅:
+- "Our preliminary data demonstrate..."
+- "We have established a robust model system..."
+- "This approach will elucidate..."
+
+**Arrogant** ❌:
+- "We are uniquely qualified..."
+- "Only our lab can do this..."
+- "This will revolutionize the field..."
+
+**Tentative** ❌:
+- "We hope to..."
+- "We will try to..."
+- "It is possible that..."
+
+### Active and Specific
+
+**Aim statements should**:
+- Start with action verbs (Determine, Identify, Elucidate, Define, Characterize, Validate, Develop)
+- Be specific and testable
+- Indicate what will be learned
+
+**Weak Aim** ❌:
+"Aim 1: Study the role of protein X in disease Y"
+
+**Strong Aim** ✅:
+"Aim 1: Determine how protein X phosphorylation regulates disease Y progression using genetic and pharmacologic approaches"
+
+### Show Feasibility
+
+**Throughout the aims page**:
+- Mention preliminary data (figures, tables)
+- Reference established methods
+- Show you have necessary resources
+- Demonstrate expertise
+- Indicate prior success
+
+**Don't**:
+- Relegate all preliminary data to Research Strategy
+- Make it seem like you're starting from scratch
+- Propose overly ambitious aims without support
+
+## Common Mistakes
+
+### Mistake 1: Too Much Background
+
+❌ Half page of background before getting to aims
+
+✅ Focused background that motivates your specific approach
+
+The aims page is NOT a mini review article. Provide only enough background to establish importance and gaps.
+
+### Mistake 2: Vague Objectives
+
+❌ "We will study the mechanisms of disease X"
+❌ "We will investigate the role of protein Y"
+
+✅ "We will identify the phosphorylation sites on protein Y that regulate its interaction with Z using mass spectrometry and mutagenesis"
+
+### Mistake 3: Overly Ambitious Scope
+
+❌ Four aims, each of which could be a separate R01
+❌ Proposing to solve multiple major questions in the field
+❌ "Boil the ocean" approach
+
+✅ Focused aims that are clearly achievable in 3-5 years
+
+### Mistake 4: Dependent Aims
+
+❌ Aim 2 and Aim 3 both require Aim 1 to succeed
+
+✅ Aims are synergistic but independent (failure of one doesn't doom the others)
+
+### Mistake 5: No Preliminary Data Mentioned
+
+❌ Seems like a fishing expedition
+❌ Reviewers wonder if it's feasible
+
+✅ Brief mentions of preliminary data throughout (refer to figures)
+
+### Mistake 6: Weak Impact Statement
+
+❌ "This will advance our understanding of X"
+❌ "Results will be published and presented"
+
+✅ "This will identify new therapeutic targets for disease X, affecting 500,000 patients annually, and provide the foundation for investigator-initiated clinical trials"
+
+### Mistake 7: Jargon-Heavy First Paragraph
+
+❌ Opening sentence full of abbreviations and specialized terminology
+❌ Assumes all reviewers are experts in your subfield
+
+✅ Opening that's comprehensible to broad scientific audience
+
+### Mistake 8: No Clear Hypothesis
+
+❌ Just listing aims without unifying framework
+❌ Purely descriptive aims
+
+✅ Clear, testable hypothesis that unifies the aims
+
+### Mistake 9: Forgetting Page Limits
+
+❌ Using 1.1 pages (will be deleted or rejected)
+❌ Tiny fonts to cram in more content (violations)
+
+✅ Exactly 1 page with compliant formatting
+
+### Mistake 10: Not Telling a Story
+
+❌ Disconnected aims that feel like 3 separate projects
+❌ No logical flow or coherence
+
+✅ Unified narrative with aims building on each other
+
+## Advanced Tips
+
+### Use Visual Elements
+
+**Figures on Specific Aims Page**:
+- NIH allows figures on aims page
+- Can be very effective to show key preliminary data
+- Must be legible (font size requirements apply)
+- Don't let figure crowd out text
+- Typical: 1 small figure or panel showing most critical data
+
+**Tables**:
+- Can summarize preliminary data compactly
+- Show patient characteristics, gene lists, etc.
+- Must be readable
+
+### Strategic Use of Bold/Italics
+
+**Appropriate**:
+- Bold aim statements to make them stand out
+- Italicize gene names (standard convention)
+- Underline key points (sparingly)
+
+**Avoid**:
+- Excessive formatting that looks cluttered
+- All caps (looks like shouting)
+- Colors (may not print/display correctly)
+
+### The "Skim Test"
+
+**Your aims page should pass the skim test**:
+- Someone reading just aim statements should understand the project
+- Bold aim statements that can be read independently
+- Each paragraph has clear topic sentence
+- Logical flow is apparent even when skimming
+
+**Exercise**: Ask colleague to read only bold/underlined text—can they understand the project?
+
+### Tailoring to Career Stage
+
+**Early Stage Investigators**:
+- Show you've thought through challenges
+- Demonstrate strong mentorship and institutional support
+- Emphasize innovation while ensuring feasibility
+- Don't over-promise
+
+**Established Investigators**:
+- Show how this extends your research program
+- Emphasize track record implicitly
+- Can propose more ambitious aims if supported by extensive preliminary data
+- Show how this opens new directions
+
+## Examples of Strong Opening Paragraphs
+
+### Example 1: Cancer Biology
+
+"Metastatic breast cancer kills over 42,000 women annually in the United States, with median survival of only 2-3 years after diagnosis. While primary tumors are often curable, metastatic disease remains incurable due to therapy resistance and tumor heterogeneity. The emergence of drug-resistant cell populations during treatment represents the major barrier to long-term survival, yet the mechanisms governing resistance evolution remain poorly understood. Understanding how tumor heterogeneity and plasticity drive resistance could reveal new therapeutic strategies to prevent or reverse treatment failure."
+
+### Example 2: Neuroscience
+
+"Alzheimer's disease (AD) affects 6.7 million Americans and is projected to reach 13 million by 2050, with annual costs exceeding $355 billion. Despite decades of research focused on amyloid-β and tau pathologies, no disease-modifying therapies exist. Emerging evidence implicates synaptic dysfunction as the earliest pathological event in AD, preceding neurodegeneration by years. The molecular mechanisms linking synaptic failure to cognitive decline represent a critical therapeutic window, yet remain poorly defined. Identifying early synaptic alterations could enable intervention before irreversible neuronal loss occurs."
+
+### Example 3: Infectious Disease
+
+"Antimicrobial-resistant (AMR) infections cause over 2.8 million illnesses and 35,000 deaths annually in the US, with healthcare costs exceeding $4.6 billion. Carbapenem-resistant Enterobacterales (CRE) represent an urgent threat, with mortality rates exceeding 50% for bloodstream infections. Despite this crisis, only two new antibiotics targeting CRE have been approved in the past decade, both with significant limitations. Novel therapeutic approaches that bypass traditional antibiotic mechanisms are urgently needed to combat this growing threat. Targeting host-pathogen interactions rather than bacterial viability represents a promising strategy to combat AMR while reducing selection pressure for resistance."
+
+## Revision Checklist
+
+Before finalizing, ensure your aims page:
+
+**Content**:
+- [ ] Opens with compelling statement of importance
+- [ ] Clearly defines the gap or problem
+- [ ] States specific, measurable objective
+- [ ] Presents testable hypothesis (or focused research questions)
+- [ ] Mentions preliminary data supporting feasibility
+- [ ] Includes 2-4 specific aims
+- [ ] Each aim is testable and achievable
+- [ ] Aims are independent but synergistic
+- [ ] Expected outcomes are clearly stated
+- [ ] Closes with impact and significance
+
+**Clarity**:
+- [ ] First paragraph is accessible to non-specialists
+- [ ] Technical terms are defined
+- [ ] Abbreviations are spelled out at first use
+- [ ] Logical flow is clear
+- [ ] Aim statements can stand alone
+- [ ] Language is confident and active
+
+**Format**:
+- [ ] Exactly 1 page
+- [ ] 0.5-inch margins
+- [ ] 11-point font or larger
+- [ ] Readable line spacing
+- [ ] Compliant with NIH formatting requirements
+- [ ] Figures (if included) are legible
+
+**Impact**:
+- [ ] Passes the "skim test"
+- [ ] Would make you excited if you were a reviewer
+- [ ] Clearly articulates significance
+- [ ] Shows feasibility without over-selling
+- [ ] Connects to health or scientific impact
+
+## Final Thoughts
+
+The Specific Aims page is where grants are won or lost. **Invest time in getting this right**:
+
+- Write 10+ drafts
+- Get feedback from colleagues and mentors
+- Test it on people outside your field
+- Read it aloud to check flow
+- Let it sit, then revise with fresh eyes
+- Study funded examples in your field
+
+**Remember**: Reviewers are reading 10-20 applications. Your aims page needs to immediately communicate importance, innovation, and feasibility—and make them want to fund your work.
+
+---
+
+**Key Takeaway**: The perfect Specific Aims page tells a compelling story in exactly one page—establishing a significant problem, presenting an innovative and feasible solution, showing preliminary evidence of success, and articulating transformative impact. Every sentence must earn its place.
+

package/lib/skills/builtin/rewrite-humanize/SKILL.md

@@ -0,0 +1,116 @@
+---
+name: rewrite-humanize
+description: Rewrite existing draft to sound more natural and less AI-templated while preserving scientific meaning, facts, numbers, citations, and claims. Use when user asks to reduce AI tone, improve flow, or humanize wording without changing conclusions.
+category: Writing & Review
+tags: [writing]
+triggers: [humanize, reduce AI tone, rewrite naturally, polish writing, improve flow, less robotic, 润色, 改写得自然, sound more natural]
+---
+
+# Rewrite Humanize
+
+## Overview
+
+Polish existing writing to reduce robotic and templated tone while keeping academic rigor and original meaning unchanged.
+
+This skill is for final language refinement after technical content is already defined. It is especially useful for computer systems and HPC papers targeting venues such as SC and HPDC.
+
+## When to Use This Skill
+
+Use this skill when the user asks to:
+- reduce AI-like tone / make writing more natural
+- improve readability, flow, and sentence rhythm
+- rewrite for publication-ready language without changing core content
+- rewrite LLM-generated drafts into native-like academic prose
+
+Do NOT use this skill to:
+- invent new results, data, citations, or claims
+- change study conclusions or interpretation direction
+- perform literature search or add external facts
+
+## Non-Negotiable Constraints (MUST)
+
+1. Preserve all factual content, numbers, units, and citations.
+2. Preserve claim strength (do not overstate or weaken conclusions).
+3. Preserve uncertainty language (`may`, `might`, `suggests`, etc.) unless user explicitly asks otherwise.
+4. Do not add new references, experiments, or unsupported statements.
+5. Keep domain terminology accurate and consistent.
+6. If source has section headings, keep the same structure unless user requests restructuring.
+
+## Role
+
+Act as a senior CS academic editor focused on natural, readable, and rigorous conference writing.
+Target style: clear technical prose suitable for top CS venues (including SC and HPDC), without sounding generic or over-polished.
+
+## Mandatory Rewrite Rules
+
+### 1) Lexical Normalization
+
+- Prefer plain, precise academic wording over inflated vocabulary.
+- Avoid overused "LLM-sounding" words unless technically required.
+- Examples:
+  - `leverage` -> `use`
+  - `delve into` -> `investigate`
+  - `tapestry` -> `context` or direct technical wording
+- Use specialized terms only when they carry necessary technical meaning.
+
+For more word-choice guidance, see `@skill/references/lexicon.md`.
+
+### 2) Structural Naturalization
+
+- For manuscript body text, convert list-like writing into coherent paragraphs.
+- Remove mechanical transitions such as `First and foremost` and `It is worth noting that`.
+- Keep logical flow through meaning, not formulaic connectors.
+- Minimize em dashes; prefer commas, clauses, or parentheses.
+
+### 3) Formatting Hygiene
+
+- In manuscript body text, do not introduce bold/italic emphasis just for style.
+- Keep LaTeX source clean; do not add unrelated formatting commands.
+- Preserve existing meaningful markup when it encodes structure or semantics.
+
+### 4) Conservative Edit Threshold (Critical)
+
+- Do not rewrite for the sake of rewriting.
+- If the input is already natural and publication-ready, keep it largely unchanged.
+- In that case, explicitly provide positive feedback in `Part 3`.
+
+## Venue Tone Notes
+
+- SC / HPDC style favors direct technical claims, reproducibility details, and restrained rhetoric.
+- Prefer concrete mechanism and evidence statements over promotional language.
+- Keep contribution framing confident but specific.
+
+For venue-specific tone patterns, see `@skill/references/cs-venue-tone.md`.
+
+## Rewrite Workflow
+
+1. Read source text and identify genuinely artificial phrasing.
+2. Classify section type (abstract, intro, methods, results, discussion, caption).
+3. Rewrite only where needed under the mandatory rules.
+4. Verify factual lock: numbers, units, equations, citations, and claims unchanged.
+5. Apply conservative threshold: revert unnecessary edits.
+6. Produce output in the required contract format.
+
+## Output Contract
+
+Return in this format:
+
+1. `Part 1: Rewritten Text`
+2. `Part 2: Integrity Check`
+- Facts/numbers/citations preserved: Yes/No
+- New claims introduced: Yes/No
+- Claim strength changed: Yes/No
+- Structural or formatting risks: brief
+3. `Part 3: Quality Assessment`
+- AI-tone issues fixed: short summary
+- If text was already strong: explicit positive feedback
+- Remaining concerns (if any): brief
+
+## References
+
+Load these as needed:
+
+- `@skill/references/lexicon.md`: word-choice normalization and phrase-level cleanup patterns.
+- `@skill/references/cs-venue-tone.md`: CS venue tone preferences for SC/HPDC-like writing.
+
+If any constraint cannot be guaranteed, state it explicitly before the rewritten text.

package/lib/skills/builtin/rewrite-humanize/references/cs-venue-tone.md

@@ -0,0 +1,57 @@
+# CS Venue Tone Notes (SC / HPDC Oriented)
+
+This reference captures lightweight tone preferences for systems and HPC papers.
+Use as guidance only; do not override user or venue template requirements.
+
+## Core Style
+
+- Lead with concrete technical problem framing.
+- State contributions in specific, testable terms.
+- Tie claims to evidence, not adjectives.
+- Keep rhetoric restrained; avoid marketing tone.
+
+## Section-Level Expectations
+
+### Abstract
+
+- One-sentence problem context, one-sentence approach, key quantitative results, and implication.
+- Avoid broad claims that are not directly supported by reported metrics.
+
+### Introduction
+
+- Clearly define bottleneck, prior limitations, and why they matter at scale.
+- Contributions should be concrete and countable.
+
+### Methods / Design
+
+- Prioritize mechanism clarity and implementation constraints.
+- Keep terminology consistent with systems/HPC practice.
+
+### Results
+
+- Report setup and metrics precisely.
+- Emphasize effect sizes and tradeoffs, not only best-case improvements.
+
+### Discussion / Limitations
+
+- Explicitly state scope boundaries and conditions where results may not transfer.
+
+## Language Patterns to Prefer
+
+- `We evaluate on ...`
+- `Compared with <baseline>, our method improves ... by ...`
+- `The gain is most visible when ...`
+- `This result suggests ... under ... conditions`
+
+## Language Patterns to Avoid
+
+- `revolutionary`, `groundbreaking`, `unparalleled` without direct evidence
+- Vague claims such as `significantly better` without metric context
+
+## Reproducibility Signal
+
+When rewriting, preserve or improve clarity around:
+- hardware/software setup
+- dataset and workload configuration
+- statistical reporting choices
+- fairness of baseline comparison

package/lib/skills/builtin/rewrite-humanize/references/lexicon.md

@@ -0,0 +1,50 @@
+# Lexicon Guidance for Humanized Scientific Writing
+
+Use this file as a preference guide, not a blind replacement table.
+Always preserve technical meaning and field-specific terminology.
+
+## Preferred Direction
+
+- Prefer plain, precise words over inflated phrasing.
+- Prefer concrete verbs over abstract nominalizations.
+- Prefer short, direct clauses over stacked qualifiers.
+
+## Common Replacements (Context-Dependent)
+
+- `leverage` -> `use`
+- `delve into` -> `investigate` or `analyze`
+- `showcase` -> `show` or `demonstrate`
+- `facilitate` -> `enable` or `support`
+- `robustly` -> `reliably` (only if evidence supports this)
+- `state-of-the-art` -> `competitive` or explicit ranking/result
+- `novel` -> specific contribution statement
+- `tapestry` -> `context` or domain-specific wording
+
+## Phrases to Avoid Unless Necessary
+
+- `First and foremost`
+- `It is worth noting that`
+- `Needless to say`
+- `In order to` (usually `to` is enough)
+- `The fact that`
+
+## Keep These Elements Stable
+
+- Terms of art (for example: `all-reduce`, `work stealing`, `NUMA`, `weak scaling`)
+- Dataset and benchmark names
+- Metric names and symbols
+- Numbers, units, equations, and citation anchors
+
+## Avoid to use following words
+
+Accentuate, Ador, Amass, Ameliorate, Amplify, Alleviate, Ascertain, Advocate, Articulate, Bear, Bolster,
+Bustling, Cherish, Conceptualize, Conjecture, Consolidate, Convey, Culminate, Decipher, Demonstrate,
+Depict, Devise, Delineate, Delve, Delve Into, Diverge, Disseminate, Elucidate, Endeavor, Engage, Enumerate,
+Envision, Enduring, Expedite, Foster, Galvanize, Harmonize, Hone, Innovate, Inscription,
+Integrate, Interpolate, Intricate, Lasting, Leverage, Manifest, Mediate, Nurture, Nuance, Nuanced, Obscure,
+Opt, Originates, Perceive, Perpetuate, Permeate, Pivotal, Ponder, Prescribe, Prevailing, Profound, Recapitulate, Reconcile, Rectify, Rekindle, Reimagine, Scrutinize, Substantiate, Tailor, Testament, Transcend, Underscore, Unveil, Vibrant
+
+## Caution
+
+Do not mechanically simplify terminology in Methods and Results sections.
+If a technical term is standard in the subfield, keep it.