Skip to content
Get packs

Guide · Agent skill

How to use the Experiment Log Standardizer agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

nature-experiment-log is an agent skill that turns lab material (photos, audio transcripts, typed notes or local files) into a standardized Markdown experiment log with YAML frontmatter. It suits researchers who keep notes as photos or voice memos and lab groups that want one naming scheme for experiments, equipment and sample batches. After this guide you will know how to install the skill, ask for a log in natural language, and check the output.

Experiment Log StandardizerYuan1z0825/nature-skills · Apache-2.0

Quick start

  1. 1Download the ZIP. Download the skill package from this directory's page.
  2. 2Install it. Unzip into ~/.claude/skills/ for all projects or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
  3. 3Give it material. Upload photos, audio transcripts or text in the session, or give a local file or folder path.
  4. 4Ask for a log. Say what you want, for example a standard experiment record, and name the output folder or ask for save-ready Markdown.
  5. 5Answer its questions. Reply to any questions about unclear temperatures, sample numbers or missing fields instead of leaving them blank.
  6. 6Check the files. Open the generated log and the archived raw folder, and confirm the IDs and frontmatter match your scheme.

What the skill makes the agent do

Once the skill is loaded, the agent takes the experiment material you give it and runs a fixed workflow: it extracts structured information using image understanding (the skill references a vision_analyze step) and text parsing, asks you about anything unclear, generates an experiment ID and a sample batch ID, writes the log, and archives the raw attachments into a dated folder.

The skill is explicit that it will not guess. If a temperature is unclear or a sample number is missing, the agent asks you rather than writing a guessed value into the log.

  • Writes one log per experiment to a path in the form {OUTPUT_ROOT}/实验日志/{体系}/{类型}/{exp_id}.md
  • Archives raw attachments to {OUTPUT_ROOT}/raw/experiments/YYYY.MM.DD_描述_EXPID/ and links them from the log
  • Updates an experiment index and appends anomaly records when the skill's templates are enabled
  • Tells you where the generated files and raw materials were saved

Install the skill (ZIP or Claude apps upload)

Download the ZIP from the skill's page on this directory. Then choose one of two routes. For Claude Code, unzip the package into ~/.claude/skills/ to make it available everywhere, or into a project's .claude/skills/ folder to make it available only in that project. In the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

The package contains SKILL.md (the instructions the agent reads), README_EN.md and README.md (English and Chinese documentation), agents/openai.yaml, manifest.yaml, three example logs in references/, and three Obsidian templates in templates/.

  • Route 1: unzip into ~/.claude/skills/ (all projects) or your project's .claude/skills/ (one project)
  • Route 2: upload in the Claude apps via Settings → Capabilities → Skills
  • Core use needs no Feishu or Obsidian installation
~/.claude/skills/nature-experiment-log/SKILL.md

Trigger it with a natural request

You do not need a special command. The skill loads automatically when you submit experiment material: uploaded images, audio or transcripts, text, a local file or folder path, or (with the optional integration) messages from a Feishu group. State what you want and what the output should be.

The repository's own example requests are a good model. Each one names the material and the goal, and the agent will ask about any missing fields.

  • “记录一个实验：316L 在 500°C 氯盐腐蚀 300 h，Ar 气氛，失重 0.0032 g。” (Record an experiment: 316L corroded in chloride salt at 500°C for 300 h, Ar atmosphere, mass loss 0.0032 g.)
  • “把这几张实验照片整理成今天的 Obsidian 实验日志。” (Turn these experiment photos into today's Obsidian experiment log.)
  • “根据这段语音转写补一条标准实验记录，并列出缺失信息。” (Build a standard record from this voice transcript and list the missing information.)

Understand the output and ID rules

Each experiment gets an ID in the form {system code}-{device code}-YYMMDD-{sequence}. The device codes in the skill are M for muffle furnace, T for tube furnace, E for electrochemical workstation, G for glovebox, F for controlled-atmosphere furnace and B for general use. System codes are yours to define, such as CL, NO, OX or HY. The sample batch ID follows {system code}-{candidate number}-B{sequence}, and the same batch ID should be kept across experiments that use the same batch so Dataview can track it.

The reference examples show the full frontmatter and body. The electrochemical example, for instance, lists fields such as exp_id, salt_batch, test_method, potential_window_V, scan_rate_mV_s and anomaly, followed by sections for purpose, steps, observations, results, anomalies and next steps. The skill notes that the YAML fields are a suggested structure you can add to or remove from.

  • Experiment ID example from the reference: OX-E-260615-001
  • Sample batch ID example from the reference: OX-A1-B001
  • Device and system codes are customizable; extend them to match your equipment
{体系代码}-{设备代码}-YYMMDD-{序号}

Use the optional Obsidian integration

The skill can write only to a normal local folder, or into an Obsidian vault. Obsidian works on local Markdown files. Adding the Dataview plugin lets the frontmatter drive automatic experiment lists, anomaly summaries and equipment usage views. To use it, you also need the obsidian skill, and you copy three templates into your vault.

If you do not name a target folder, the agent returns save-ready Markdown first and does not choose a path for you.

  • Copy templates/experiment-index.md to 实验日志/实验索引.md (Dataview dashboard)
  • Copy templates/anomaly-log.md to 实验日志/异常记录.md (anomaly records)
  • Copy templates/equipment-tracking.md to 实验日志/公共/设备与试剂追踪.md (equipment and reagent tracking)

Optional Feishu input and safety notes

To pull material from a Feishu group, you use the separate feishu-cli-integration skill and lark-cli. The bot must be added to the target group and have the im:message, im:message.group_msg and im:resource permissions. Group IDs and bot setup are handled by that skill's configuration. Note that --output accepts only relative paths, so change into the raw/experiments/ archive folder first.

Safety: this skill runs no scripts, but it does use the network when the Feishu integration is enabled, and that integration reads group messages and attachments. If your material is confidential or regulated, review where files are sent before enabling it. Local-folder and typed-input use does not require the Feishu setup.

  • Runs scripts: no
  • Uses network: yes, through the optional Feishu integration
  • Pull messages: lark-cli im +chat-messages-list --chat-id oc_*** --page-size 30 --sort asc
  • Download images: lark-cli im +messages-resources-download --message-id *** --file-key *** --type image --output <relative path>
lark-cli im +chat-messages-list --chat-id oc_*** --page-size 30 --sort asc

Tips for better logs

Give the agent everything you have in one pass: the date, sample, conditions, process steps, observations and attachments. Missing items come back as questions, so a complete first message saves a round trip.

Customize before you rely on it. The folder names and codes reflect a specific lab's conventions, so set your own system codes and device codes, and create subfolders under wiki/实验日志/{体系}/ for your experiment types. Check the three reference examples to see the expected level of detail.

  • Name your sample batches consistently so sample_batch stays the same across experiments
  • Answer the agent's questions about temperatures and sample numbers instead of skipping them
  • Keep your ID scheme written down so the team uses the same codes

What people use it for

  • Turning a day's photos of furnace and electrochemical work into one standardized log per experiment
  • Converting voice-memo transcripts into structured records and a list of missing details
  • Keeping a shared sample batch ID across several corrosion or stability experiments for Dataview tracking
  • Maintaining an Obsidian experiment index, anomaly log and equipment tracking page for a lab group

Frequently asked questions

Is the nature-experiment-log skill free to use?

The repository for this skill lists the Apache-2.0 licence, but the SKILL.md frontmatter states MIT, so check the licence files in the repository before you redistribute it. The skill page here does not list a price. Confirm any cost of the agent platform itself with that platform.

What do I need to run nature-experiment-log?

You need an agent with file read and write access to your chosen output folder, and image understanding, since the skill uses a vision step. Obsidian with the Dataview plugin and the obsidian skill are optional, as are feishu-cli-integration and lark-cli for Feishu group input.

Does the skill run scripts or use the internet?

According to the listing, it runs no scripts and does use the network. The network use comes from the optional Feishu integration, which reads group messages and attachments, so review that before using confidential data.

How do I install nature-experiment-log in Claude?

Download the ZIP from this page and unzip it into ~/.claude/skills/ for all projects or into a project's .claude/skills/ folder. In the Claude apps you can instead upload it under Settings → Capabilities → Skills.

Why does the agent keep asking me questions instead of writing the log?

The skill is designed not to guess experimental conditions or results. If a temperature is unclear or a sample number is missing, it asks you to confirm before writing that field.

Can I use this skill without Obsidian, and how does it compare to other tools?

Yes. The core flow writes Markdown to a normal local folder, and Obsidian is optional. The skill is built around specific Chinese-language naming and one lab's equipment codes, so it is a poor fit if you need a generic tool without customizing codes. The source does not compare it with other tools.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use this skill if you keep lab notes as photos or voice memos and want every experiment written in one consistent Markdown format with traceable IDs. Pick something else if you need a ready-made, general-purpose tool, or if you handle confidential data and are not willing to review what the optional Feishu integration sends over the network.

Download the skill