Agent Skills: Tilt Up

Tilt Up

Principles (Always Active)

These apply whenever working with Tiltfiles, Tilt errors, or dev environment bootstrap:

Fix the Tiltfile, Not the Symptoms

• Fix the source config directly - Tiltfile, Dockerfile, k8s manifest, or helm values
• Never add shell workarounds - no wrapper scripts, no || true, no try/except pass
• Never hard-code ports, paths, hostnames, image tags, or container names that should be dynamic
• Never add fallbacks that mask the real error - if a resource fails, the failure must be visible
• Never add sleep/retry loops for flaky dependencies - fix dependency ordering via resource_deps() or k8s_resource(deps=)
• Never add polling for readiness that Tilt already handles - use k8s_resource(readiness_probe=) or probe configs

Express Dependencies Declaratively

• Port conflicts: fix the port allocation source, don't pick a different port
• Resource ordering: use resource_deps(), not sequential startup scripts
• Env vars: use silo.toml or gen-env output, not inline defaults
• Image availability: use image_deps or deps, not sleep-until-ready

Tilt Live-Reloads

After editing a Tiltfile, Tilt picks up changes automatically. Never restart tilt up for:

• Tiltfile edits
• Source code changes
• Kubernetes manifest updates

Restart only for: Tilt version upgrades, port/host config changes, crashes, cluster context switches.

Workflow (When Explicitly Starting Tilt)

Step 1: Assess Current State

1. Check if tilt is already running:

   PROJECT=$(basename "$(git rev-parse --show-toplevel 2>/dev/null)" || basename "$PWD")
   zmx list --short 2>/dev/null | grep -q "^${PROJECT}-tilt$"
   

   If running, check health via tilt get uiresources -o json and skip to Step 3.

2. Check for required env files (.localnet.env, .env.local, silo.toml):

   • If silo.toml exists, use silo up path
   • If gen-env script exists, run it first
   • If neither, check project README for bootstrap instructions

3. Check for k3d cluster or Docker prerequisites.

Step 2: Start Tilt in zmx

Follow the zmx skill patterns:

PROJECT=$(basename "$(git rev-parse --show-toplevel 2>/dev/null)" || basename "$PWD")
SESSION="${PROJECT}-tilt"

if zmx list --short 2>/dev/null | grep -q "^${SESSION}$"; then
  echo "Tilt session already exists: $SESSION"
else
  zmx run "$SESSION" 'tilt up'
  echo "Started tilt in zmx session: $SESSION"
fi

For silo projects: silo up instead of tilt up.

Step 3: Monitor Bootstrap

Poll for convergence:

1. Wait 10s for initial resource registration
2. Poll every 15s, up to 20 iterations. Include docker-compose container health (composeResourceInfo.healthStatus) — an Up (unhealthy) compose container keeps runtimeStatus=ok and is otherwise invisible, so bootstrap can look "done" while canton/splice/postgres are silently failing their HEALTHCHECK:

   tilt get uiresources -o json | jq -r '.items[] | select(.status.runtimeStatus == "error" or .status.updateStatus == "error" or .status.updateStatus == "pending" or .status.composeResourceInfo.healthStatus == "unhealthy") | "\(.metadata.name): runtime=\(.status.runtimeStatus) update=\(.status.updateStatus) compose=\(.status.composeResourceInfo.healthStatus // "-")"'
   

3. Track resources: pending -> in_progress -> ok
4. Success: all resources reach runtime=ok, update=ok (or not_applicable) AND no docker-compose resource is composeResourceInfo.healthStatus == "unhealthy"
5. If resources stabilize in error, OR a compose resource stays unhealthy, proceed to Step 4. For an unhealthy compose probe, read the real cause with docker inspect <compose-project>-<svc> --format '{{json .State.Health}}' — often the mounted healthcheck script calls a CLI the image lacks (the service is up; fix the probe script, don't disable the check)

Step 4: Diagnose and Fix Errors

For each resource in error state:

1. Read logs: tilt logs <resource> --since 2m
2. Read the Tiltfile and relevant k8s manifests
3. Identify root cause in the config (not the running process)
4. Apply fix following the Principles above
5. Tilt live-reloads - re-poll status to verify

After 3 fix iterations on the same resource without progress:

• Report the error with full logs
• Identify whether it's a Tiltfile bug, upstream dependency, or infrastructure problem
• Do not silently skip or disable the resource

Step 5: Report

## Tilt Status: <healthy|degraded|errored>

**Resources**: X/Y ok
**Session**: zmx $SESSION

### Errors (if any)
- <resource>: <root cause> — <what was fixed or what remains>