Yuan1z0825/nature-skills / skills/nature-writing
nature-writing
VerifiedDraft or restructure scientific manuscript arguments, sections, and initial-submission materials from author-provided evidence. Use for 论文写作、章节起草、论证重构、正文压缩、首次投稿材料. Use nature-polishing for language-only edits to existing prose and nature-response for post-decision correspondence.
static/fragments/section/experiments.md
View on GitHub# Section: Experiments / Results (writing)
## Default evidence ladder
`establish the phenomenon -> stress-test it -> rule out alternatives -> broaden it -> interpret it -> bound it`
Each subsection has a claim-first opening, then data support.
Depending on the paper, instantiate this as a discovery loop, a core-capability
plus validation envelope, or a capability ladder. Load
`../../../../nature-shared/core/nature-results-discussion.md` for the three
archetypes and the same-level-repetition test.
## Drafting rules
- Load `../../../../nature-shared/core/main-text-discipline.md` before deciding
which analyses enter the main text. Classify results by their function in the
paper and build the shortest sufficient evidence chain.
- Stay mainly in past tense.
- Report what was observed, under what conditions, with what quantitative support.
- Use statistics correctly and sparingly. Every test needs a stated hypothesis.
- Keep core discovery and necessary support in the main text. Route robustness,
non-central heterogeneity, provenance detail, alternative inference, and edge
cases to SI unless they change the central interpretation.
- **Each major claim needs adequate evidence across the manuscript and SI.** Do
not force every comparison, ablation, or stress test into the main text; if
adequate evidence is absent from the full record, mark it for follow-up rather
than drafting around it.
- Normally report the descriptive quantity and primary inferential statistic in
the main text. Put secondary inference and diagnostics in SI unless required
or conclusion-changing.
## Results syntax (vs Discussion)
Results sentences usually report:
- `was detected` / `increased` / `showed` / `enabled` / `achieved`
Close each coherent evidence unit with the bounded inference it supports.
Calibrated interpretation (`suggests`, `indicates`, `likely because`) may remain
in Results when it directly answers the current experiment and the supporting
evidence is visible there. Move literature synthesis, broad implications, and
extended mechanistic reconciliation to Discussion.
## Common failure modes when drafting
- Mixing observation with an interpretation that is not supported by the
current experiment, or allowing local interpretation to expand into a broad
Discussion.
- Citing supplementary data when the result should be in the main text.
- Appending robustness or reviewer-defense prose until the central evidence chain
disappears.
- Repeating the same effects, intervals, and P values in Results and captions.
- Vague comparisons (`higher than control`) without effect size, sample size, or test.
- Per-paragraph claims without per-paragraph evidence.
- Reusing an earlier baseline or control as the centre of a later subsection
and restating the same claim, instead of making the new perturbation,
falsification, stronger comparator, or boundary the decisive evidence.
## Deeper reference
For ML/conference-style experiment sections — baselines, ablations, metrics, tables, figures — open `references/experiments.md`.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |