PuPu: Skill for Claude Code

PuPu Test API — Skill

A local HTTP REST endpoint on PuPu (dev mode only, bound to 127.0.0.1) that lets you drive PuPu like a human: create chats, switch models, send messages, read state, take screenshots. Use it to verify your code changes actually behave correctly in the running app.

When to use

• After making code changes that touch chat creation, message send, model selection, toolkits, characters, or any chat-related path — run a quick regression to confirm the happy path still works
• When debugging "the UI is wrong" — pull /v1/debug/state to see what the app thinks vs. what it shows
• When debugging "why didn't X happen" — pull /v1/debug/logs to see renderer console + main stdout (Flask logs are Phase 2)
• When you need to demonstrate a fix to the user with a real screenshot

Pre-flight check

Before invoking the API, confirm PuPu is running in dev mode:

ls "$HOME/Library/Application Support/pupu/test-api-port" && cat "$HOME/Library/Application Support/pupu/test-api-port"

Expected: a JSON {port, pid, started_at}. If the file is missing, ask the user to run npm start in the PuPu repo. If the file exists but pid references a dead process, ask the user to restart PuPu.

How to call

Quick path: curl

PORT=$(node -e "console.log(JSON.parse(require('fs').readFileSync(process.env.HOME + '/Library/Application Support/pupu/test-api-port')).port)")
BASE="http://127.0.0.1:$PORT/v1"

# Always retry 503 not_ready a couple times — it just means renderer is still initializing
for i in 1 2 3; do
  S=$(curl -s -o /dev/null -w "%{http_code}" "$BASE/debug/state")
  [ "$S" = "200" ] && break
  sleep 0.3
done

Better path: use the helper

scripts/test-api/client.mjs does port discovery + 503 retry for you:

node -e "import('./scripts/test-api/client.mjs').then(({client}) => client.GET('/debug/state').then(s => console.log(JSON.stringify(s, null, 2))))"

For a complete end-to-end smoke (create chat → send → screenshot → cleanup), run the bundled script: