testbench-mqtt
Skill235 starsChanged 3 months agoMITView on GitHub
Use this skill whenever tests involve MQTT communication — starting/stopping the mosquitto broker on the testbench Pi, publishing test messages, subscribing to topics, or verifying ESP32 MQTT client behavior. The broker runs on the Pi's WiFi AP network (192.168.4.1:1883) so DUTs can reach it without internet. Use for MQTT integration tests, pub/sub verification, and broker lifecycle management. Triggers on "MQTT", "broker", "mosquitto", "publish", "subscribe", "topic", "MQTT test".
- Sends data out
In this repository17 other files it gives its agents.
- buildSkill
- commissionSkill
- defineSkill
- esp-idf-handlingSkill
- esp-pio-handlingSkill
- grill-meSkill
- harnessSkill
- sdr-receiverSkill
- setup-actionSkill
- signal-generatorSkill
- testbench-bleSkill
- testbench-debugSkill
- testbench-installSkill
- testbench-integrationSkill
- testbench-loggingSkill
- testbench-test-handlingSkill
- testbench-wifiSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 56
Ensure device is on testbench WiFi (see testbench-wifi)
- Distinctive: line 63
Start broker, let device connect
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
ESP32 MQTT Broker
Base URL: $TESTBENCH_URL — see Step 0
Step 0: Point at a bench
There are several benches and their addresses move, so nothing here writes one
down. $BENCH is not usable either — a container cannot resolve mDNS.
Discover the bench and export its URL:
export TESTBENCH_URL=$(sudo python3 .claude/skills/esp-idf-handling/discover-testbench.py \
--url --name <bench-hostname>)
curl -s "$TESTBENCH_URL/api/info" # confirm before anything else--url refuses to guess when more than one bench answers, so --name is
required whenever a second bench is powered on. TESTBENCH_URL is the same
variable pytest --wt-url falls back to.
The testbench can run an MQTT broker (mosquitto) for testing ESP32 devices that use MQTT for communication. The broker is accessible to devices connected to the testbench's WiFi AP.
Endpoints
Request and response shapes: FSD Appendix D.12.
Examples
# Start the MQTT broker
curl -X POST $TESTBENCH_URL/api/mqtt/start
# Check broker status
curl $TESTBENCH_URL/api/mqtt/status
# Stop the MQTT broker
curl -X POST $TESTBENCH_URL/api/mqtt/stopMQTT Broker Details
| Property | Value |
| Broker | the bench address (from the LAN) or 192.168.4.1 (from the testbench AP) |
| Default port | 1883 |
| Authentication | None (open broker for testing) |
Common Workflows
- Test ESP32 MQTT client:
- Ensure device is on testbench WiFi (see testbench-wifi)
POST /api/mqtt/start— start broker- Device connects to
192.168.4.1:1883 - Monitor device behavior via serial or UDP logs (see testbench-logging)
POST /api/mqtt/stop— stop broker when done
Test MQTT disconnect/reconnect:
- Start broker, let device connect
POST /api/mqtt/stop— device loses MQTT- Monitor device's reconnection behavior
POST /api/mqtt/start— device should reconnect
On a shared broker, don't stop it — isolate the one device. If the broker also carries other devices (or the device under test uses a production broker over a tunnel rather than the bench mosquitto), stopping it is collateral damage. Cut MQTT for a single device with a source-scoped firewall rule on the broker host instead (needs root there):
# black hole (timeout): device sees silence, its keepalive eventually fires the will iptables -I INPUT -s <device-ip> -p tcp --dport 1883 -j DROP # connection refused (TCP reset): device sees an immediate refusal iptables -I INPUT -s <device-ip> -p tcp --dport 1883 -j REJECT --reject-with tcp-resetThe two are different tests:
DROPis the "broker went away" black hole that exercises the last-will keepalive timeout;REJECTis the "broker actively refuses" path. Delete the rule (-Dinstead of-I) to restore. Prefer this over/api/mqtt/stopwhenever the broker is not exclusively the DUT's.- Test MQTT + WiFi together:
- Start AP + broker
POST /api/wifi/ap_stop— device loses both WiFi and MQTTPOST /api/wifi/ap_start+POST /api/mqtt/start— restore both- Verify device recovers
A last-will topic is death-only — don't test for it "returning to online"
The MQTT Last Will is published by the broker, on the device's behalf, only
when the connection drops uncleanly. A device typically registers a retained
will like <id>/LWT = offline and signals liveness on a separate birth or
status topic (e.g. <id>/STATUS {"online":true}, republished on every connect
and periodically). Nothing ever writes LWT = online — the will topic stays
offline from the last drop until the next one.
So when checking recovery after an outage, watch for the birth/status message
to be republished, not for the will topic to flip back. And because the
retained will persists, a stale offline cannot be told apart from a freshly
fired one by reading the retained value — to prove a will fired, seed a
sentinel first (mosquitto_pub -t <id>/LWT -r -m sentinel), then confirm the
broker overwrites it with offline once the keepalive expires.
A retained status read mid-outage is a stale corpse
mosquitto_sub -C 1 on a retained status/telemetry topic returns the last
published value, which during (and right after) an outage is the copy from
before the device dropped — its counters are frozen. Reading it to judge
recovery gives a false negative: e.g. mqtt_disconnects still shows the
pre-outage number because the device has not reconnected and republished yet.
Judge recovery by the device resuming live publishing, not by the retained value. Subscribe for longer than the publish cadence and require ≥2 arrivals (the immediate retained copy, then at least one fresh publish); only then read counters. An MQTT reconnect does not reset uptime, so "small uptime" cannot stand in for "fresh" — only a live arrival can.
Don't over-specify long black-hole durations
A spec that says "cut MQTT for 30 minutes" is usually testing "the device never restarts on MQTT loss." Unlike WiFi (which often has a restart-after-N-minutes rule), MQTT loss typically has no restart timer at all — so a few minutes proves "never restarts" exactly as well as 30, and the extra 25 just confirm the same nothing. Reserve long windows for behaviour actually gated on a long timer (a WiFi restart rule, a WireGuard rekey/handshake expiry). Also: a telemetry-drop counter can only grow if the device is publishing telemetry that overflows — with no sensor/meter attached there is no stream, so that sub-criterion is N/A on a bare bench and needs a data source (a fake-data build or real hardware).
Troubleshooting
| Problem | Fix |
| Broker won't start | Check if mosquitto is installed on the testbench Pi |
| Device can't connect | Ensure device is on testbench WiFi; use broker IP 192.168.4.1 from AP clients |
| Broker status shows stopped | Start it with POST /api/mqtt/start |
| Device "never comes back online" | You are probably watching the /LWT will topic — it is death-only; watch the birth/status topic instead (see above) |
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

