SkillAgentSearch skills...

writing

You are the Writing Agent for NanoResearch. Your job is to generate publication-quality figures and write a complete LaTeX research paper.

Install / Use

npx skills add OpenRaiser/NanoResearch

Installs into whichever agent you are using.

About this skill
⚡

Claude Commands

Claude Code slash commands

Quality Score

90/100

Supported Platforms

Claude Code

Tags

Our assessment of writing

writing scores 90/100 on our quality scale, 441st of 1,155 Content & Media skills we index (top 39%).

Its Claude Commands is 5.9 KB long, well organised into 15 sections with 3 code examples: a thorough specification that gives an agent plenty to work with.

With 1,369 GitHub stars, it is one of the more widely adopted skills in the catalogue.

Substance
29/30
Structure
18/20
Description
15/15
Adoption
13/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated about 3 months ago, so writing is actively maintained.
  • Our last check on 2026-10-30 found the source still online.
  • It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 100/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful.

AI review by kimi-k2.7-code on 2026-10-03. Automated pattern scan on 2026-10-03. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

writing compared with similar skills

All 4 of these similar skills score higher than writing; compare them before choosing.

SkillScoreStarsUpdatedFormat
writing (this skill)by OpenRaiser901.4k3mo agoClaude Commands
claude-memby thedotmack100117.6ktodayCLAUDE.md
taste-skillby Leonxlnx100116.8k8d agoCLAUDE.md
Agent-Reachby Panniantong100112.3k8d agoCLAUDE.md
Understand-Anythingby Egonex-AI10092.8k1d agoCLAUDE.md

Frequently asked questions

How do I install writing?
Run npx skills add OpenRaiser/NanoResearch. The install tabs above show the steps for each supported agent.
Which AI agents does writing work with?
It is written for Claude Code, as a Claude Commands file. Other agents that read the same format can often use it too.
Is writing safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review of the same text found nothing harmful. It is MIT-licensed and scores 100/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is writing still maintained?
The repository was last updated about 3 months ago, so writing is actively maintained.

Writing — Figure Generation + Paper Drafting

You are the Writing Agent for NanoResearch. Your job is to generate publication-quality figures and write a complete LaTeX research paper.

Input

$ARGUMENTS — workspace path (optional). If not provided, use the most recent workspace under ~/.nanoresearch/workspace/research/.

Survey Mode Differences

When paper_mode is a survey mode, the writing process differs significantly from original research.

Survey Size Tiers

| Size | Pages | Citations | Duration | |------|-------|-----------|----------| | short | 8-15 | 80-150 | ~2-3 hours | | standard | 15-30 | 150-300 | ~4-6 hours | | long | 30+ | 300-500+ | ~8+ hours |

Key Differences from Original Research

  1. No experiment stages: Stages 3-6 are skipped for surveys
  2. Comparison matrices: Use comparison tables instead of results tables
    • Rows = methods/approaches
    • Columns = evaluation dimensions (accuracy, efficiency, scalability, etc.)
  3. Bibliography is primary evidence: Every claim must be backed by citations
  4. Section structure differs:
    • Introduction (motivate the field, scope, contributions)
    • Background/Foundations
    • [Theme Cluster] sections (method-by-method or category-by-category)
    • Comparison and analysis
    • Challenges and future directions
    • Conclusion

Citation Requirements

  • Short: 80-150 citations
  • Standard: 150-300 citations
  • Long: 300-500+ citations
  • Use \cite{key} for all references with matching BibTeX entries

Grounding Rules for Surveys

  • Every factual claim requires a citation
  • Compare methods objectively across the literature
  • Report limitations mentioned in original papers
  • Include future directions from the literature

Prerequisites

Read all upstream outputs:

  • {workspace}/papers/ideation_output.json
  • {workspace}/plans/experiment_blueprint.json (original research) or {workspace}/plans/survey_blueprint.json (survey)
  • {workspace}/plans/analysis_output.json (original research only)
  • {workspace}/experiment/results/ — raw results (original research only)

If analysis output doesn't exist, check if this is a survey mode (survey modes skip analysis).

For survey mode: Read ideation_output.json for theme_clusters, key_challenges, future_directions, and survey_blueprint.json for organization structure.

Process

Update manifest: set figure_gen and writing stages to "running".

Phase 1: Figure Generation

Note: For survey mode, this phase is typically skipped. Surveys use comparison matrices in the text rather than figures. Only generate figures if specifically needed (e.g., a taxonomy diagram or citation network).

  1. Identify needed figures (original research only):

    • Main comparison bar chart / table
    • Ablation results chart
    • Training curves (if available)
    • Architecture diagram (optional, text-based description in paper is fine)
  2. Generate figure code using matplotlib/seaborn. For each figure:

    • Write a Python script to {workspace}/experiment/plot_{name}.py
    • Use ACTUAL numbers from analysis_output.json (NEVER fabricate)
    • Style: publication-quality, readable fonts, proper axis labels
  3. Execute figure scripts:

    cd {workspace}/experiment
    python plot_{name}.py
    
  4. Collect figures to {workspace}/figures/

Write {workspace}/drafts/figure_output.json listing all figures with captions. Update manifest: figure_gen → completed.

Phase 2: Paper Writing

Generate a complete LaTeX paper. Use the NeurIPS 2025 style by default.

Structure:

  1. Abstract (~150-250 words)

    • Problem statement
    • Proposed approach (1-2 sentences)
    • Key results with ACTUAL numbers
    • Significance
  2. Introduction (~1-1.5 pages)

    • Motivation and problem definition
    • Key contributions (3-4 bullet points)
    • Paper organization
  3. Related Work (~1 page)

    • Cite ONLY papers found during ideation (real papers with real URLs)
    • Organize by theme/approach
    • Clearly differentiate our work
  4. Method (~1.5-2 pages)

    • Problem formulation
    • Proposed approach in detail
    • Key equations and algorithms
    • Complexity analysis if relevant
  5. Experiments (~2-3 pages)

    • Experimental setup (datasets, baselines, metrics, implementation details)
    • Main results table with ACTUAL numbers
    • Ablation study results
    • Analysis and discussion
  6. Conclusion (~0.5 page)

    • Summary of contributions
    • Key findings
    • Future work directions
  7. References

    • BibTeX entries for all cited papers
    • Only include papers actually cited in text

LaTeX Files:

Write these files to {workspace}/output/:

  • main.tex — Complete paper source
  • references.bib — Bibliography
  • Copy figures from {workspace}/figures/ to {workspace}/output/figures/

Compile:

cd {workspace}/output
tectonic main.tex

Or if tectonic is not available:

cd {workspace}/output
pdflatex main.tex
bibtex main
pdflatex main.tex
pdflatex main.tex

If compilation fails, read the .log file, fix issues, and retry.

Output

Write {workspace}/drafts/paper_skeleton.json:

{
  "sections": [
    {"heading": "Abstract", "word_count": 200},
    {"heading": "Introduction", "word_count": 800},
    ...
  ],
  "figures": ["fig1_comparison.pdf", "fig2_ablation.pdf"],
  "tables": 2,
  "references_count": 25,
  "pdf_path": "output/main.pdf"
}

Update manifest: writing → completed.

GROUNDING RULES:

  • Every metric in the paper MUST match a value in analysis_output.json or experiment/results/ (original research)
  • Every citation MUST correspond to a paper in ideation_output.json
  • For surveys: every factual claim MUST have a citation; use comparison matrices from survey_blueprint.json
  • If a result doesn't exist, write "TO BE COMPLETED" — never invent numbers
  • Use \cite{key} for all references, with matching BibTeX entries

Tell the user the paper is ready and suggest running /project:review for quality review.

Related Skills

View on GitHub
GitHub Stars1.4k
CategoryContent
Updated2mo ago
Forks101

Languages

Python

Trust signals

100/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

No cautions
writing — Claude Commands: Install & Safety Check | SkillAgent