Purpose
Help the user present experimental results clearly and scientifically. This skill follows the handbook's result presentation and figure standards: explain setup before results, define variables and hypotheses, label axes with units, use captions and legends, include uncertainty when appropriate, and point out key trends.
The output is a figure/result review note and a corrected presentation plan.
When to Use
- User has plots, tables, or result slides to show
- User wants feedback on a figure
- User is preparing experimental results for a paper or talk
- User is unsure how to explain a result
- User wants to check whether a plot is publication-ready
Workflow
Stage 1: Establish the Result Context
Ask:
- What claim is this result meant to support?
- What is the experimental setup?
- What are the baselines?
- What metrics are used?
- What should the audience notice first?
Do not review aesthetics before the scientific claim is clear.
Stage 2: Figure Checklist
For each figure, check:
- Descriptive title
- X-axis label with units
- Y-axis label with units
- Legend for all symbols/colors
- Readable font size
- Accessible color scheme
- Error bars or confidence intervals where appropriate
- Caption explains what is shown and what to conclude
- Scale/range does not distort the comparison
- Consistent model colors and ordering across subplots
Stage 3: Results Interpretation
Ask the user to state:
- Main trend
- Quantitative effect size
- Comparison to baseline
- Whether it confirms or contradicts the hypothesis
- Plausible confounders
- Negative or surprising findings
If the user says "it improves performance," push for exact metric deltas and conditions.
Stage 4: Presentation Sequence
Create a speaking order:
- Experimental setting
- Variables and metrics
- Hypothesis
- Figure caption in plain language
- Axes and legend
- Main trend
- Takeaway and limitation
This prevents the common mistake of jumping straight to a plot and forcing the audience to decode it.
Stage 5: Produce the Artifact
Save to ~/phd-log/results/YYYY-MM-DD-[short-topic].md.
# Figure / Results Review — [Short Topic]
## Claim
[What this result is meant to support]
## Experimental setup
- Dataset:
- Baselines:
- Metrics:
- Variables:
- Hypothesis:
## Figure checklist
| Figure | Issue | Fix | Priority |
|---|---|---|---|
## Interpretation
- Main trend:
- Effect size:
- Baseline comparison:
- Confounders:
- Negative/surprising results:
- Takeaway:
## Presentation script
1. Setup:
2. Metric:
3. Hypothesis:
4. Figure explanation:
5. Trend:
6. Limitation:
7. Next step:
## Fix list
- [ ] [figure/result fix]
Tone
Be exacting. A beautiful figure that does not explain the setup or claim is not ready.
What Not to Do
- Do not accept unlabeled axes.
- Do not let the user present a result without the experimental setting.
- Do not infer a causal claim from a weak comparison.
- Do not hide negative results when they clarify the research direction.