Clinical Tempo · ClawHub context skill

Clinical Tempo is the reference HealthTech Protocol superapp: React hub + dedicated routes, Node/Express API, Tempo settlement, MPP / x402 machine payments. This skill tells agents where context lives and how to avoid known traps (see CLAWHUB.md).

Quick reference

When to activate (triggers)

Use this skill automatically when:

1. The user @-mentions llm-full.txt, CLAWHUB, Clinical Tempo, MPP, dance-extras, or Tempo testnet/mainnet.
2. The task touches server/index.js, server/payments.js, or src/danceExtras*.ts.
3. Docs are edited that appear in scripts/build-llm-full.mjs (bundle sources).
4. The user uploads ClawHub / OpenClaw / Copilot context questions.

Recommended: Cursor / IDE

1. @ public/llm-full.txt for broad changes; @ CLAWHUB.md when debugging past incidents.
2. Project rules: repo root AGENTS.md or .cursor/rules if present — align with README.md.

See references/openclaw-clinical-tempo.md for OpenClaw workspace file hints.

Installation

From this repository (authoritative):

# Skill lives at:
.cursor/skills/clawhub/

Manual copy to OpenClaw skills dir (optional):

cp -r .cursor/skills/clawhub ~/.openclaw/skills/clinicaltempo-clawhub

ClawHub (publish): Zip clawhub/ (this folder) so the listing includes SKILL.md, references/, assets/, hooks/, scripts/. See README.md in this folder for a file manifest.

OpenClaw: optional Anyway plugin

This skill is context-only; @anyway-sh/anyway-openclaw extends the OpenClaw runtime with additional capabilities (separate from Clinical Tempo). Install when you want both:

openclaw plugins install @anyway-sh/anyway-openclaw
# If your setup requires it:
openclaw gateway restart

Pair with: this skill (or ClawHub listing) + optional clinicaltempo-clawhub hook. Details: references/openclaw-clinical-tempo.md.

Repository map (where to look)

clinicaltempo/
├── public/llm-full.txt          # Generated — do not hand-edit; run npm run build:llm
├── CLAWHUB.md                   # Tribal knowledge: successes, failures, checklists
├── README.md                    # Routes, stack, quick start
├── HEALTHTECH_USE_CASES.md       # Flow-by-flow API contract
├── server/index.js              # Express routes, integrations, MPP proxies
├── server/openapi.mjs           # OpenAPI 3.1 for GET /openapi.json (MPPScan)
├── server/payments.js           # Chain IDs, charge helpers
├── src/hubRoutes.ts             # Hub directory of all /routes
├── src/danceExtrasLiveMpp.ts    # Browser MPP helpers (live flows)
├── src/danceExtrasJudgeWire.ts  # Judge-score wire snippets
├── .github/copilot-instructions.md
└── scripts/build-llm-full.mjs   # Source list for llm-full.txt

First load (full orientation)

1. Prefer public/llm-full.txt (or /llm-full.txt from a running build) — includes CLAWHUB.md in the bundle.
2. Regenerate after doc edits: npm run build:llm.

Bundle sources

Exact list: assets/LLM-BUNDLE-SOURCES.md (mirrors build-llm-full.mjs).

EVVM: llm-full.txt vs llms-full.txt

Debugging (tribal knowledge)

Read CLAWHUB.md for:

• What succeeded / failed (purl, AgentMail, 402 loops, stale Express on 8787, etc.)
• Repeatable checks (e.g. GET /api/dance-extras/live)

Deeper: references/troubleshooting.md.

Key implementation pointers

Concrete snippets: references/examples.md.

Copilot Chat integration

GitHub Copilot does not load this folder automatically. Options:

1. Commit /.github/copilot-instructions.md (already in Clinical Tempo repo).
2. Paste from references/copilot-and-agents.md into chat or org instructions.

Quick prompts:

• “Use llm-full.txt as context for this PR.”
• “Scan CLAWHUB.md for 8787 / MPP / purl before changing the server.”
• “After this task, suggest one CLAWHUB Success or Failure line (no secrets).”
• “Regenerate public/llm-full.txt — which files are inputs?” → assets/LLM-BUNDLE-SOURCES.md

Promotion: CLAWHUB vs llm-full

Verification script

From repo root (optional):

./.cursor/skills/clawhub/scripts/verify-clinical-tempo-context.sh

Checks that public/llm-full.txt exists and reminds you of the live MPP GET check.

OpenClaw hook (optional)

Parity with self-improving-agent: injects a virtual CLINICAL_TEMPO_CONTEXT_REMINDER.md on agent:bootstrap (skips sub-agents). No network I/O.

# From this skill folder (or repo: .cursor/skills/clawhub/)
cp -r hooks/openclaw ~/.openclaw/hooks/clinicaltempo-clawhub

openclaw hooks enable clinicaltempo-clawhub

• hooks/openclaw/HOOK.md — metadata + enable/disable.
• hooks/openclaw/handler.js — CommonJS handler (primary).
• hooks/openclaw/handler.ts — TypeScript handler for OpenClaw (types from openclaw/hooks at runtime).

Full notes: references/openclaw-clinical-tempo.md.

Best practices

1. Never paste real .env secrets into prompts — use .env.example names only.
2. After editing any file listed in build-llm-full.mjs, run npm run build:llm before claiming “docs are updated in the bundle.”
3. Prefer GET /api/dance-extras/live over guessing whether the server is new.
4. For EVVM, answer no to optional Sepolia registry registration until policy allows (see docs/EVVM_TEMPO.md).

Operations manual (same rigor as self-improving-agent)

This skill mirrors that package’s shape: quick reference, OpenClaw integration, optional hooks/scripts, templates, and promotion — but maps “project memory” to CLAWHUB.md + README.md instead of only .learnings/.

Promotion targets (where tribal knowledge goes)

Entry templates: assets/templates/CLAWHUB-ENTRY.md.

Optional: .learnings/ format (LRN / ERR / FEAT)

If you maintain a parallel log (e.g. in OpenClaw workspace), use the same ID patterns as self-improving-agent (LRN-YYYYMMDD-XXX, etc.). Starters ship under assets/learnings/. Authoritative for this repo remains CLAWHUB.md.

Periodic review (before a large change)

# Grep CLAWHUB for an area
rg -n "MPP|402|8787|TIP-20|purl" CLAWHUB.md

# Confirm bundle is fresh after doc edits
npm run build:llm

# Smoke API (server must run on 8787)
curl -sS http://127.0.0.1:8787/api/dance-extras/live | head -c 400

Detection triggers (load this skill when)

• User mentions llm-full.txt, CLAWHUB, Clinical Tempo, HealthTech, Tempo, MPP, 402, NHS, TIP-20, 8787, ClawHub, or OpenClaw.
• Editing server/index.js, scripts/build-llm-full.mjs, or src/hubRoutes.ts.
• Preparing a zip for clawhub.ai or debugging a published install.

Claude Code / Codex hooks (optional)

Configure .claude/settings.json or .codex/settings.json — full paths in references/hooks-setup.md.

Multi-agent matrix

Skill extraction (new package from a fix)

When a CLAWHUB entry is reusable across repos, consider a standalone skill:

./.cursor/skills/clawhub/scripts/extract-skill.sh my-org-skill --dry-run
./.cursor/skills/clawhub/scripts/extract-skill.sh my-org-skill

Then replace the scaffold under .cursor/skills/my-org-skill/ using assets/SKILL-TEMPLATE.md.

Files in this package

See also

• Published skill: clawhub.ai/arunnadarasa/clinicaltempo — browse ClawHub for versioned skills
• Ecosystem synergy (mpp-nanogpt-modal, nanochat, OpenClaw): docs/ECOSYSTEM_SYNERGY.md
• Structural inspiration: self-improving-agent (references, assets, scripts, hooks, .learnings patterns)
• Clinical Tempo repo: README.md, CLAWHUB.md