How does this skill handle failed experiments compared to a standard AI prompt?

This skill uses a strict evidence ledger system that prevents the AI from smoothing over data or ignoring failed trials, ensuring every precise metric and 'negative' result is preserved for your PI's review.

Which platforms and file types are compatible with the Lab Meeting Report skill?

The skill works with local Markdown, PDF, CSV, and code files, and can export finished reports directly to Notion, Feishu/Lark, or presentation formats like Marp and Quarto.

What exactly is included in the purchase of this skill?

Upon purchase, you receive the production-ready skill file, a configuration guide for cloud integrations (Notion/Feishu), and templates for various report modes like Research Progress and Journal Club.

Can I manually edit the report before it gets published to Notion or converted into slides?

While the skill handles the heavy lifting of synthesis, it generates a local Markdown 'source of truth' first, allowing you to manually refine any technical nuances before final publishing.

Does this skill support summarizing multiple research papers for a Journal Club?

Yes, the skill includes a specific 'Paper Review' mode designed for Journal Clubs that synthesizes multiple PDFs into structured notes with evidence grounding.

Lab Meeting Report

Weekly Lab Meeting Report: Microscopy Segmentation

Synthetic demonstration. All project details and results are fictional.
Reporting period: July 6–12, 2026 | Pre-meeting report

Weekly Overview

Progress: Completed three seeds with the current U-Net configuration. Added stain normalization and fixed the data loader's random state. All three runs completed training.
Sources: weekly-notes.md, results.csv

Key evidence: Dice scores were 0.812, 0.806, and 0.809 for seed-11, seed-23, and seed-37 respectively.
Source: results.csv

Discussion needed: Physical batch size 4 caused an out-of-memory failure on a 16 GB GPU despite AMP and cache clearing. Decide whether to borrow a GPU with at least 24 GB or use gradient accumulation.
Sources: weekly-notes.md, results.csv

Previous Action Review

Action	Status	Artifact
Complete three seeds using the current configuration	Complete	`results.csv`

Sources: weekly-notes.md, results.csv

Experimental Results

Configuration	Seed	Batch size	Dice	IoU	Peak GPU memory	Status
Current	seed-11	2	0.812	0.691	14.8 GB	Complete
Current	seed-23	2	0.806	0.683	14.9 GB	Complete
Current	seed-37	2	0.809	0.687	14.8 GB	Complete

Source: results.csv

These observations confirm that all three runs completed. The supplied material does not establish statistically significant improvement.

Failed Attempts and Evidence Limits

Boundary dropout: Dice was 0.774 for seed-23, compared with 0.806 for the current configuration using the same seed. This week's evidence does not support retaining this branch.

Physical batch size 4: Training failed at the first optimizer step on the 16 GB GPU. AMP and cache clearing did not resolve the failure.

Sources: weekly-notes.md, results.csv

Literature idea: The boundary-aware auxiliary loss in the paper notes was studied on fluorescence microscopy. Its applicability to the current bright-field data is unknown, and it has not been implemented in this project.

Source: paper-notes.md

Blocker and Requested Decision

The current GPU cannot complete training with physical batch size 4.

Choose between

Borrowing a GPU with at least 24 GB to test physical batch size 4.
Keeping physical batch size 2 and using gradient accumulation for effective batch size 4.

Source: weekly-notes.md

Next Actions

Action	Deliverable	Completion criterion
Run the batch configuration experiment using the agreed route	Training logs and a comparison table for three seeds	All runs finish without OOM; record Dice, IoU, and peak GPU memory
Assess the boundary-aware auxiliary loss	A documented keep-or-defer decision	Use the current configuration as comparator; do not treat paper results as results on this dataset

If no 24 GB GPU is available, proceed with gradient accumulation. No performance improvement threshold has been predefined.

Sources: weekly-notes.md, paper-notes.md

Sources

weekly-notes.md
results.csv
paper-notes.md

About this skill

Prepare your next lab meeting from the research material you already have.

Turn weekly notes, papers, experiment results, figures, and code changes into a concise Markdown report covering progress, evidence, blockers, and next steps.

Create weekly research summaries, progress reports, and journal club notes.
Preserve failed experiments, negative results, and uncertainty.
Keep facts, interpretations, and hypotheses distinct.
Avoid invented metrics, citations, and explanations.
Generate reports in Chinese or English.
Use optional guidance for presentation preparation and Feishu/Lark or Notion publishing.

Getting started

“Use lab-meeting-report to read my weekly notes, results, and paper notes, then prepare a report for a 10-minute lab meeting.”

Open source under the MIT license.

Frequently Asked Questions

Lab Meeting Report