d-laub/dlaub-togo

scientific-writing-draft

Use when outline.md is approved and you need full manuscript prose in the user's voice. Stage 3 of scientific-writing — produces draft.md from the outline, then enforces the style guide in a separate final pass.

First seen Jun 17, 2026

Installation

$ npx skills add d-laub/dlaub-togo --skill scientific-writing-draft

Similar popular skills

Related neighbors and high-traction skills in the same topics — useful to compare before installing.

Also in this package

Other skills from d-laub/dlaub-togo · top by installs.

npx skills add d-laub/dlaub-togo

Browse all from d-laub/dlaub-togo

More details

Agent compatibility

Declared targets from SKILL.md / docs. Unmarked agents are not listed — the skill may still install via the CLI.

Claude Code Not declared
Cursor Not declared
Codex Not declared
GitHub Copilot Not declared
Windsurf Not declared
Gemini CLI Not declared
Cline Not declared
OpenCode Not declared

Repository health

Default branch main
Open issues 0
Status Active

Skill metadata

Parsed from SKILL.md frontmatter.

More metadata
type
technique

Package contents

Files included with this skill beyond the listing page.

  • skill md SKILL.md 4,949 B
  • docs SUMMARY.md 245 B

History

  1. First seen on skills.sh
  2. First recorded snapshot · 21 installs

SKILL.md

Scientific Writing — Stage 3: Draft

Overview

Turn the approved outline.md into full prose, then enforce the style guide in a separate pass. Two phases:

  1. Draft — section-by-section prose from the outline, in the user's voice. Content fidelity is the priority.
  2. Style correction — a dedicated subagent re-reads the full style guide with fresh context and corrects the whole draft.

Why split them: during a long drafting conversation the style guide drifts out of context and its rules stop being applied. A final subagent whose only context is the style guide + the draft applies the guide cleanly.

Every sentence traces to an outline bullet — drafting does not introduce new claims.

Method backbone — Steps 13–15

Drafting is Steps 13–15 of the fifteen steps: turn each planned paragraph (its Step-11 purpose statement) into one paragraph of three parts (detail: scientific-writing/fifteen-steps.md):

  • Topic sentence (Step 13). Often the paragraph's purpose statement itself; or a question, a turning point ("Having established X, we turn to Y"), a complication ("…but it fails to account for…"), or a development ("This can be made concrete with a model"). If the paragraph doesn't support the paper's main point, rewrite or cut it.
  • Supporting sentences (Step 15). Typically 3–6. For Methods/Results/Discussion responses, lift them from the outline's method/result/response lists. Equations, algorithms, and code count as supporting sentences. Strike any sentence that doesn't advance the paragraph's purpose.
  • Concluding sentence (Step 14). Closes the theme the topic sentence opened (a declaration → develops/summarizes it; a question → answers it) and segues to the next paragraph in logic if not syntax. If the topic and concluding sentences are disconnected, rewrite both.

This is the phrasing stage; the thinking is already fixed in the outline. Style polish is deferred entirely to Phase 2.

Load context first

Read: story.md, outline.md, references.md, and the selected style guide. Default style: styles/dlaub-style.md (this skill's directory). If multiple styles exist in styles/, ask the user which to use. Record the chosen guide's path — Phase 2 needs it.

Phase 1 — Draft (per section)

  1. Write prose for the section from its outline.md bullets, one paragraph per Step-11 purpose statement using the topic→supporting→concluding contract above, in the voice of the selected style guide.
  2. Fidelity review — every claim still traces to an outline.md bullet and its backing pointer/references.md entry. No new unbacked claims, no dropped figure callouts, no altered numbers.
  3. Next section.

Write draft.md in the story dir. Do not run the full style checklist here — that is Phase 2's job. A rough-but-faithful draft is the correct Phase 1 output.

Phase 2 — Style correction (dedicated subagent)

After the complete draft exists, dispatch ONE subagent whose only context is the style guide and the draft. This is required, not optional: do not hand-polish style inline in this conversation, because the guide has already drifted out of context here.

Dispatch a subagent with this task:

Read <path-to-selected-style-guide> in full, then read <story-dir>/draft.md.
Apply every rule in the style guide to the entire draft, working its end-of-guide checklist paragraph by paragraph.
Make style-only edits, in place in draft.md:
- Do NOT add, remove, or alter any claim.
- Do NOT change any number, citation marker, or figure callout (text or position).
- Do NOT change section structure or headings.
Return a short list of the categories of change you made (e.g. "removed 6 em-dashes, de-hedged 4 sentences, fixed 2 repeated figure callouts").

When the subagent returns, skim its change list and spot-check that no numbers or callouts moved.

Output

  • draft.md with citation markers and figure callouts preserved exactly as in the outline.
  • After Phase 2: a short handoff noting any spots where the outline was thin and prose had to stretch, plus the subagent's style-change summary.

Red flags — STOP

  • A sentence with a claim not in outline.md → remove it or send the claim back to Stage 2.
  • A number changed from the outline → revert; outline numbers are the source of truth.
  • Skipping the Phase 2 subagent and hand-styling inline → the style guide has drifted out of context; dispatch the subagent.
  • The style subagent changed a number, citation, or figure callout → revert that edit; the pass is style-only.

Done

Hand draft.md to the user for review.