mushanghai-longevity

mushanghai-longevity

muShanghai Longevity Month 2026 — a community self-experimentation program running in Shanghai in May–June 2026. Two parallel cohorts (Longevity Month Experiment + BioBloom Hackathon), three shared research tracks (Decoding TCM, Lifestyle, Proof of Living — the Longevity Month edition of muShanghai Build in Public), one shared canon. This repo holds the role definitions, lifecycle docs, practical Shanghai resources, and the AI-agent skills (for Claude Code + OpenCode) that walk participants through proposing, running, logging, demoing, and retrospecting their experiments.

Disclaimer. Everything in this repo is for education, exploration, and self-observation — not medical diagnosis or treatment. Do not paste identifiable health data, other people's data, or details of medical conditions into chat. High-risk interventions (anything beyond lifestyle / breath / movement / off-the-shelf wearables) need Track Leader and Month-organizer approval. Full safety canon: skills/mushanghai-canon/references/safety.md.

Table of Contents

• TL;DR
• Who this repo is for
• Getting started on May 12 (90 minutes)
• Repo structure
• Cohorts (May 2026)
• The Three Tracks (shared across both cohorts)
• AI-Agent Skills
  • Current skills (May 2026 cycle MVP)
  • Install
  • Local skills/ vs global ~/.agents/skills/
  • Use
  • Distribution status
• Documentation Conventions
• License

TL;DR

Pick your role → install the skills → talk to Claude Code (or OpenCode). The skills handle everything from "I have a vague hypothesis" through "I just demoed at OpenFlow Bazaar."

• New attendee on May 12? → jump to Getting started on May 12.
• Already know your role? → use the router in Who this repo is for.
• Just want to read the docs? → start with the three role definitions: experiment-leader/ROLE.md, track-leader/ROLE.md, participants/ROLE.md.

Who this repo is for

Track-Leader note. The program is not recruiting more Track Leaders for the May 2026 cycle. track-leader/ROLE.md is a "what we offer" doc for ELs and Participants, not a recruitment page.

Getting started on May 12 (90 minutes)

If you're attending the May 12 AI-Empowered Onboarding session, follow this exact sequence. The whole thing fits in the scheduled 90 minutes.

1. Clone or download this repo.

git clone https://github.com/frankji-groundcontrol/mushanghai-longevity.git
cd mushanghai-longevity

2. Install the muShanghai skills (live-updating symlink path recommended — you'll get repo updates automatically with git pull):

mkdir -p ~/.claude/skills ~/.agents/skills
for s in mushanghai-canon mushanghai-onboarding mushanghai-experiment-leader \
         mushanghai-track-leader mushanghai-participant-log \
         mushanghai-demo-prep mushanghai-retrospective; do
  ln -sfn "$(pwd)/skills/$s" ~/.claude/skills/$s
  ln -sfn "$(pwd)/skills/$s" ~/.agents/skills/$s
done

3. Open Claude Code or OpenCode. If you don't have either yet, that's fine — mushanghai-onboarding will walk you through installing them.

4. Type this in a new session:

"Help me get started with muShanghai."

The mushanghai-onboarding skill auto-loads and runs a five-section flow:

Facilitators: see skills/mushanghai-onboarding/assets/may-12-onboarding-script.md for the 90-minute time-blocked agenda, common blockers, and success criteria.

Repo structure

Cohorts (May 2026)

• Longevity Month Experiment — May 19 – June 5. Experiments run 5 to 14 days (14-day protocol max). Demo Day June 5.
• BioBloom Hackathon — May 17 – May 22. Recruitment Night May 17, hackathon officially starts May 18 (no scheduled events), unified proposal deadline May 19, build May 19–22, Demo Day May 22 at OpenFlow Bazaar.

Both cohorts share the unified May 19 proposal submission deadline and the same three tracks. Full calendar: skills/mushanghai-canon/references/may-2026-calendar.md.

The Three Tracks (shared across both cohorts)

1. Decoding TCM — Frank + Buddy (Yale PhDs)
2. Lifestyle — Xiao He + Rebecca
3. Proof of Living (Longevity Month edition of muShanghai Build in Public) — Track Leaders: Natalia + Mariam; Program-wide Build in Public leads: Sunny + Tang Ming. Open enrollment by hashtag — #muShanghai #BuildinPublic #muShanghaiLongevity #ProofofLiving. Full spec: skills/mushanghai-canon/references/proof-of-living.md.

Track scopes + core questions: skills/mushanghai-canon/references/three-tracks.md.

AI-Agent Skills

This repo ships AI-agent skills for use with Claude Code / OpenCode. Skills live in skills/.

Current skills (May 2026 cycle MVP)

Install

Pick ONE of:

# Static copy — does NOT auto-update when this repo changes
for s in mushanghai-canon mushanghai-onboarding mushanghai-experiment-leader \
         mushanghai-track-leader mushanghai-participant-log \
         mushanghai-demo-prep mushanghai-retrospective; do
  cp -r skills/$s ~/.claude/skills/
  cp -r skills/$s ~/.agents/skills/
done

# Symlink — live-updates as the repo changes (RECOMMENDED for the May 2026 cycle)
for s in mushanghai-canon mushanghai-onboarding mushanghai-experiment-leader \
         mushanghai-track-leader mushanghai-participant-log \
         mushanghai-demo-prep mushanghai-retrospective; do
  ln -sfn "$(pwd)/skills/$s" ~/.claude/skills/$s
  ln -sfn "$(pwd)/skills/$s" ~/.agents/skills/$s
done

First-time users: start by invoking mushanghai-onboarding (Step 4 of Getting started on May 12). It installs the upstream skill packs, sets up Tavily + Firecrawl MCPs, and walks you through everything else.

Verify install:

ls ~/.claude/skills/ | grep mushanghai
ls ~/.agents/skills/ | grep mushanghai

Local skills/ vs global ~/.agents/skills/

This repo's skills/ folder is the source of truth for the muShanghai skills. There is no npm package and no npx skills add distribution path yet — that's deferred to post-May-22. To use a skill in Claude Code or OpenCode, you have to make it visible to the client by placing it (or a symlink to it) in the client's global skill directory:

cp -r vs ln -sfn — pick deliberately:

• cp -r copies the skill once. If we update the skill in this repo, your copy stays stale until you re-run cp -r. Use this if you want a frozen version (e.g., you've forked the skills and don't want upstream changes).
• ln -sfn symlinks the skill. If we update the skill in this repo (just git pull inside your local clone), your install updates automatically the next time Claude Code reads the skill. Use this unless you have a reason not to — the May 2026 cycle is moving fast and you'll want updates without re-running the install loop.

If you're on macOS and npx skills add (used by the upstream obra/superpowers and mattpocock/skills packs) ever places skills in ~/.agents/skills/ but not ~/.claude/skills/, there's a one-time symlink workaround inside skills/mushanghai-onboarding/references/superpowers-install.md.

Use

In Claude Code or OpenCode, after install:

Help me draft my experiment proposal for muShanghai

…or any phrase containing "draft proposal", "Four Musts check", "review this proposal", "BioBloom proposal", "seed hypothesis", or "Track Leader review".

The skill auto-loads and starts in router mode. It asks whether you're drafting (Mode A) or reviewing (Mode B).

Distribution status

• Source: committed to this repo at skills/
• .skill packaging: deferred until post-May-12 onboarding
• npx skills add distribution: deferred until post-May-22

Documentation Conventions

Apply these to every doc in this repo, including any new files added later:

• English only. No Chinese in document body (canonical Chinese names live in source-citation footers only when unavoidable).
• Table of Contents. Every file with more than ~3 H2 headings gets a ## Table of Contents block placed right after the intro quote and before the first H2 (typically ## TL;DR). Use H2/H3 depth; skip H4. GitHub-style anchor slugs (lowercase, hyphenated, punctuation stripped — em-dashes and & / produce double hyphens).
• TL;DR first. Every role/spec doc opens with a ## TL;DR immediately after the TOC.
• Two Parallel Cohorts framing. Any doc that touches May 2026 program structure must surface the two parallel cohorts (Longevity Month Experiment + BioBloom Hackathon) prominently — not buried mid-doc.
• 14-day protocol max. No references to 4-week, ~17-day, ~2.5-week, or month-long experiment durations.
• Three unified tracks — Decoding TCM, Lifestyle, Proof of Living. No 4-track BioBloom structure, no "Everyday Health / Frontier Life Sciences / Stories of All Things" remnants.
• Four Musts (not Three): Problem, Monitoring/Observation, Intervention/Interaction, Deliverables. The "Intervention/Interaction" item explicitly carves out creative-leaning work (Proof of Living) where Interaction substitutes for a formal scientific Intervention.
• Source citations at the bottom of long docs, listed as italic single-line refs to canonical sources (TCM-Track-Guidebook, Track General Template, Hackathon Protocol Template, Luma event pages, etc.).

License

Mozilla Public License 2.0.