La Le Me

"La Le Me" is a bilingual Chinese-English GitHub Skill published according to the Agent Skills directory conventions: it turns a low-burden, once-a-day bowel movement log into a gentle review of personal rhythms, dietary cues, and body care. It responds to the "Body, Sleep, and Health Management" bounty under the "Safety and Recovery" direction by Goushi, aiming to help the body gradually achieve a stable rhythm and true rest, rather than resorting to remedial actions after overdrawing. The repository uses skills/la-le-me as the sole canonical source and can be deployed to host environments that support Skills, such as Codex, GitHub Copilot, and Claude, according to host conventions.

It is not a diagnostic tool and does not replace doctors, pharmacists, or registered dietitians. When encountering red flags such as bleeding, black stools, severe abdominal pain, vomiting, inability to pass gas, severe dehydration, or high fever, the Skill will suspend routine recipes and conditioning advice, prioritizing medical triage. After providing immediate action steps, users can choose to authorize a one-time use of their approximate location to find nearby emergency rooms, proctology departments, colorectal surgery, or gastrointestinal surgery; the search ensures no delay in emergency care and does not require an exact residential address.

Design Philosophy

The project draws inspiration from the low-burden daily greetings requested by the user, akin to the "Are You Dead Yet" (死了么) app style, as well as the structured logging, trend observation, and permission control concepts of the Apple iOS Health App; it does not copy their code, interfaces, icons, copy, or other protected materials. This project has no affiliation or authorization relationship with Apple, the aforementioned app, or any other brands the name might bring to mind.

What It Can Do

• Record form, frequency, time of day, effort, pain, and feelings of incomplete evacuation using the user's own words or the Bristol 1-7 scale.
• Only reports food and changes as verifiable temporal associations, without claiming single events as causality or allergies.
• Summarizes personal trends based on 2-4 weeks of records; missing dates remain unknown, without treating "not recorded" as "no bowel movement".
• Generates a one-day recipe plan considering region, season, allergens, dietary preferences, and water restriction requirements.
• Under red flags, first gives action levels, then searches for nearby emergency rooms based on the user's one-time authorization; for non-emergencies requiring an in-person visit, it can search for nearby proctology or colorectal-related departments.
• Subject to host support and explicit user choice, sets daily reminders configured by time, time zone, frequency, quiet hours, and copywriting, supporting pause, resume, modification, or deletion.
• Reminders default to neutral copy that does not expose health topics and a text-only (no image) mode; if the user actively selects and the host supports media attachments, cute patterns in flat friendly, pixel, hand-drawn line art, or bright sticker styles are optionally available.
• Validates and aggregates local records using an offline Python script, defaulting to no internet connection, no write-back, and no persistent storage.
• Accepts Simplified Chinese, Traditional Chinese, English, and mixed Chinese-English input; can output Chinese, English, or paragraph-by-paragraph bilingual text. Red flags, limitations, and location authorization prompts in both languages must be equivalent.

Main Language Interface / Language

When invoking the Skill for the first time without having posed a health question, two basic language options will be displayed:

拉了么 / La Le Me
请选择语言 / Choose language
1. 中文
2. English

Hosts supporting quick options can display these as two selection controls; other hosts use numbered text. After selection, it enters the session interface of the corresponding language. If paragraph-by-paragraph bilingual output is needed, users can continue to type "中英双语" or bilingual output. If the initial message already contains a health question in a clear language, it will be answered directly; red flags will not wait for language selection. The Skill itself does not provide an independent graphical interface; whether host menus, buttons, and notification permissions are localized is determined by the host.

Installation

skills/la-le-me is the canonical Skill directory of this repository; upon release, .github/skills/la-le-me and .claude/skills/la-le-me should be generated by scripts and committed with the repository as native copies:

Install from a public GitHub repository to an Agent supported by skills.sh:

npx skills add BulletsHo/la-le-me --skill la-le-me

After updating the canonical Skill, use the following commands to rebuild and check the two native copies:

python scripts/sync_skill_distributions.py
python scripts/sync_skill_distributions.py --check

The currently publicly reported basic upload format for Red Skill is a single Markdown file, so upload RED-SKILL.md; do not use the repository ZIP as a platform runtime dependency. This single file already contains the safety, location authorization, and reminder degradation rules that must still be adhered to even without references/, scripts, or image attachments; see RED-SKILL.md for complete information and source notes. Platform review, account eligibility, and available permissions are subject to the current pages of the Xiaohongshu Creator Service Platform.

After installation, you can directly type in the conversation:

Use $la-le-me to log my bowel movement today, and tell me which most useful piece of information is still missing.

Or, without saving data first, you can directly ask:

Use $la-le-me. Over the last week, I've mostly had Bristol type 1-2, with straining, but no abdominal pain, vomiting, or bleeding. Please give me a minimal, sustainable adjustment suggestion.

Reminder Patterns

Patterns are turned off by default and only displayed when explicitly chosen by the user and the host supports media attachments. The four sets of SVGs and homologous transparent PNGs are all original works of this repository and contain no external assets or web references.

Data and Security

By default, it only processes information actively provided by the user in the current conversation; it does not read or guess history, does not ask for poop photos, and does not request exact addresses or identity information. Only when the user explicitly selects local data mode is skills/la-le-me/scripts/poop_log.py used. Seeking nearby medical care only uses the minimum location necessary for the search after completing triage and obtaining one-time authorization; location is not written into the bowel movement log. Reminders only retain the time, time zone, frequency, quiet hours, copy, and pattern preferences confirmed by the user, without embedding symptoms, history, or location.

Medical safety boundaries and sources are located at:

• Medical Safety and Triage
• Bristol Stool Form Guide
• Diet and Seasonal Recipes
• Privacy, Data, and Automation
• Location Consent and Nearby Medical Care
• Daily Reminders and Patterns

Validation

Run in the repository root directory:

python scripts/sync_skill_distributions.py --check
python -m unittest discover -s tests -v
python skills/la-le-me/scripts/poop_log.py validate --input tests/fixtures/synthetic-records.json
python skills/la-le-me/scripts/poop_log.py summarize --input tests/fixtures/synthetic-records.json

For complete validation methods, synthetic safety use cases, and known limitations, see VALIDATION.md.

Platform directory conventions refer to: Agent Skills Specification, GitHub Copilot Agent Skills, Claude Skills, Vercel Agent Skills, and [Red Skill Upload Instructions]（https://xhslink.cn/o/17UVMkggQks）. True host capabilities users subject to the current documentation and permissions of each platform.

License

Code and original documentation are released under the MIT License. Official Bristol flashcards are not copied with the repository; this repository only uses self-rewritten brief text and source links.