verify

name	verify
description	How to verify ari-hooks changes end-to-end against a real Claude Code session

Verifying ari-hooks

The surface is a Claude Code session in a folder where the hooks are installed. Unit tests (npm test) exercise the hook binary against a stub HTTP server, but real verification means booting claude and watching the hooks fire.

Recipe

1. Stub the Ari API — a small node http.createServer on 127.0.0.1:<port> serving the endpoints the hooks call (/agent-activities POST, /agent-tasks GET). Log method/url/auth.
2. Isolated config — point ARI_HOOKS_HOME at a temp dir containing config.json: { "token": "test", "apiUrl": "http://127.0.0.1:<port>" }.
3. Demo project — temp dir with .claude/settings.json whose hook commands use the absolute local bin: node /Users/maxheckel/Sites/ari-hooks/bin/ari-hooks.js hook <event>.
4. Headless (context checks) — cd demo && ARI_HOOKS_HOME=... claude -p "...". Hooks run in -p mode; ask Claude about injected context to confirm additionalContext landed.
5. TUI (display checks) — expect stalls (claude waits on terminal capability queries a dumb pty never answers). Use tmux: tmux -L verify new-session -d -x 100 -y 40 -c <demo> "env ARI_HOOKS_HOME=... claude", sleep ~6-8s, tmux -L verify capture-pane -p. Drive with send-keys. First boot in a new folder shows a trust dialog (send Enter).
6. Failure probes — API down / garbage 200 response: session must boot clean; failures land in $ARI_HOOKS_HOME/error.log.

Clean up: tmux -L verify kill-server, kill stub server pids.