ts-paper-write

Drafts all LaTeX body sections of a Traitement du Signal proposal paper from a blueprint and bibliography.

1.1k|19|Updated Jun 18, 2026
One-click install
npx skills add https://github.com/Spark-To-Paper-Skills/spark-to-paper-skills --skill ts-paper-write-spark-to-paper-skills
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ts-paper-write
Source: https://github.com/Spark-To-Paper-Skills/spark-to-paper-skills/tree/main/skills/ts-paper-write
Command: npx skills add https://github.com/Spark-To-Paper-Skills/spark-to-paper-skills --skill ts-paper-write-spark-to-paper-skills

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Writing a full academic paper draft section-by-section often produces inconsistent terminology, fabricated results, and LaTeX that fails to compile. This Skill drafts every section of a Traitement du Signal (TS) proposal paper in one holistic pass, enforcing strict per-section recipes, a no-fabricated-results rule, and IIETA citation style. ## Core Features & Use Cases - One-pass section drafting: Generates sections/<id>.tex body files plus abstract.tex from a blueprint and refs.bib, following per-section recipes (abstract, introduction, related work, method, experiments, analysis, conclusion) with word bands and structure rules. - Honesty enforcement: Prohibits fabricated metrics in proposal mode and supports a data-aware mode where every reported number must come from results.facts.json; the bundled draft_lint.py script machine-checks these rules plus AI-tell phrases, non-ASCII characters, and template shape contracts. - Figure and table scaffolding: Emits FIGURE-SPEC placeholders for a later figure stage, result tables with blank -- cells, notation tables, and algorithmicx pseudocode blocks. - Use Case: Given a research blueprint and a curated BibTeX file, produce a complete, lint-clean LaTeX draft of a TS proposal paper ready for the refine stage. ## Quick Start Draft all sections of my TS proposal paper from blueprint.json and refs.bib in the workdir, then run draft_lint.py until it passes.

Frequently Asked Questions about ts-paper-write

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I draft a full LaTeX academic paper from a blueprint?

Provide a blueprint.json and refs.bib in the workdir, then run the write stage to generate one body file per section under sections/ plus abstract.tex. Each section follows a recipe with word bands, subsection counts, and citation rules, and draft_lint.py validates the result.

How does the skill prevent fabricated results in a proposal paper?

In proposal mode, draft_lint.py hard-fails any prose containing percentages, decimal scores, signed deltas, multipliers, word-form magnitudes, or placeholder tokens like [X]%. Result table cells must stay blank as `--`, and only forward-looking hedged language is allowed.

What is data-aware mode and when should I use it?

Data-aware mode activates when template.json sets results_mode to data_aware, meaning real experimental results exist. Every number written must appear in results.facts.json, prose switches to past tense, and the author fills result tables with real measured values instead of blanks.

Does the skill support citation styles other than numeric?

Yes, citation style is template-driven. If template.json sets citations.style to author_year, the draft uses \citep and \citet commands; for numeric style such as TS it uses plain \cite with bibkeys from refs.bib.

Why does draft_lint.py fail my draft even though the prose looks fine?

Common causes are word counts outside the template band, wrong subsection or contribution-item counts, missing notation table references, non-ASCII characters outside math, bold text in prose, or AI-tell phrases like "it is worth noting". Run the linter and fix every reported violation until it exits zero.

What are the limitations of the automated linting checks?

The linter cannot judge citation-claim match, design rationale quality, hedging tone, terminology consistency, or prose flow. These judgment items require a manual self-review pass, and fabricated bare-integer results are not caught by the number audit.