Agent not responding | Friends of the Crustacean 🦞🤝 | Page 1

meager flint Apr 29, 2026, 12:17 PM

#

Skills with missing dependencies
1password, apple-notes, apple-reminders +36 more
2 cron jobs failed
Work services status monitor, OpenClaw issue 73838 status monitor

{"0":"{"subsystem":"diagnostic"}","1":"lane wait exceeded: lane=session:agent:main:main waitedMs=2801 queueAhead=0","_meta":{"runtime":"node","runtimeVersion":"22.22.1","hostname":"unknown","name":"{"subsystem":"diagnostic"}","parentNames":["openclaw"],"date":"2026-04-29T12:15:03.247Z","logLevelId":4,"logLevelName":"WARN","path":{"fullFilePath":"file:///home/claude/.npm-global/lib/node_modules/openclaw/dist/subsystem-rHhUC6qs.js:147:14","fileName":"subsystem-rHhUC6qs.js","fileNameWithLine":"subsystem-rHhUC6qs.js:147","fileColumn":"14","fileLine":"147","filePath":".npm-global/lib/node_modules/openclaw/dist/subsystem-rHhUC6qs.js","filePathWithLine":".npm-global/lib/node_modules/openclaw/dist/subsystem-rHhUC6qs.js:147","method":"logToFile"}},"time":"2026-04-29T08:15:03.247-04:00","hostname":"claude-A7-Max","message":"lane wait exceeded: lane=session:agent:main:main waitedMs=2801 queueAhead=0"}
{"0":"{"subsystem":"gateway/ws"}","1":"⇄ res ✓ skills.status 54ms conn=29bbaf7a…3957 id=2a39834c…0c94","_meta":{"runtime":"node","runtimeVersion":"22.22.1","hostname":"unknown","name":"{"subsystem":"gateway/ws"}","parentNames":["openclaw"],"date":"2026-04-29T12:15:10.891Z","logLevelId":3,"logLevelName":"INFO","path":{"fullFilePath":"file:///home/claude/.npm-global/lib/node_modules/openclaw/dist/subsystem-rHhUC6qs.js:147:14","fileName":"subsystem-rHhUC6qs.js","fileNameWithLine":"subsystem-rHhUC6qs.js:147","fileColumn":"14","fileLine":"147","filePath":".npm-global/lib/node_modules/openclaw/dist/subsystem-rHhUC6qs.js","filePathWithLine":".npm-global/lib/node_modules/openclaw/dist/subsystem-rHhUC6qs.js:147","method":"logToFile"}},"time":"2026-04-29T08:15:10.891-04:00","hostname":"claude-A7-Max","message":"⇄ res ✓ skills.status 54ms conn=29bbaf7a…3957 id=2a39834c…0c94","traceId":"6a9e922637aeb5ff2ae2ff22b7455b57","spanId":"f4a16972095bf378","traceFlags":"01"}

primal coveBOT Apr 29, 2026, 12:17 PM

#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

woeful treeBOT Apr 29, 2026, 12:18 PM

#

That looks like a few separate signals, not one single root cause:

“Skills with missing dependencies” is usually informational. It means optional skills like 1password, apple-notes, etc. aren’t usable until their deps are installed. It won’t stop the agent unless the run/cron specifically needs one of those skills.
“2 cron jobs failed” is the thing to inspect. Please paste:

openclaw --version
openclaw status --all
openclaw cron status
openclaw cron list
openclaw cron runs --id <jobId> --limit 20

For each failed run, also check:

#

openclaw tasks show <runId-or-sessionKey>
openclaw logs --follow

lane wait exceeded ... waitedMs=2801 queueAhead=0 means the session queue waited ~2.8s before starting that agent run. By itself that’s just a warning, not necessarily fatal. If it repeats with long waits, something may be occupying the main session lane.

If those cron jobs are using session main, I’d consider moving recurring monitors to --session isolated once we confirm the failure, so they don’t block normal chat replies.

Docs: https://docs.openclaw.ai/help/faq and https://docs.openclaw.ai/gateway/troubleshooting#cron-and-heartbeat-delivery

#

I’m in a fresh session and rebuilt what I could from the workspace:

no session-resume-*.json file
no data/highlights-main.jsonl
recent changes were mostly docs under openclaw-repo/

So I don’t have prior memory beyond what you pasted here.

From your logs, two quick reads:

“Skills with missing dependencies” usually just means optional skills aren’t eligible on this host. Seeing things like apple-notes / apple-reminders on Linux is normal and usually not the reason the agent stopped replying.
lane wait exceeded ... waitedMs=2801 is a queue-delay warning, not a crash by itself.

The simplest next step is:

openclaw --version

#

openclaw status --all

Paste those, and if you can, also include:

openclaw logs --follow

then send one test message to the agent and grab the resulting log lines.

If you want, I can help you read the outputs line-by-line after that.

meager flint Apr 29, 2026, 12:22 PM

#

it was working yesterday

woeful treeBOT Apr 29, 2026, 12:23 PM

#

Yeah — if it worked yesterday (April 28) and failed today (April 29), I’d focus on what changed overnight, not the missing skills list.

The useful signal is the failed cron run history. The missing skill deps are probably noise unless those monitor jobs explicitly use one of those skills.

Please paste these:

openclaw --version
openclaw cron list
openclaw cron runs --id <jobId> --limit 20

Run the last command for both failed jobs.

Likely buckets:

gateway restarted / machine slept / cron missed its window
model/provider/auth failed during the monitor run

#

delivery target failed, so the monitor ran but didn’t post
one of the jobs is using session main and got tangled with the main agent lane

That lane wait exceeded ... waitedMs=2801 is only ~2.8s, so by itself I wouldn’t treat it as the cause. The failed cron run details should tell us the actual error.

meager flint Apr 29, 2026, 12:23 PM

#

{
"ts": 1777408220623,
"jobId": "2eee5f9c-4664-4c41-948d-1939655a9a87",
"action": "finished",
"status": "ok",
"runAtMs": 1777408200005,
"durationMs": 20616,
"nextRunAtMs": 1777409100000,
"model": "gpt-5.5",
"provider": "openai-codex",
"usage": {
"input_tokens": 784,
"output_tokens": 62,
"total_tokens": 40768
},
"delivered": false,
"deliveryStatus": "not-delivered",
"delivery": {
"intended": {
"channel": "last",
"to": null,
"source": "last"
},
"resolved": {
"ok": false,
"to": null,
"source": "last",
"error": "Channel is required (no configured channels detected). Set delivery.channel explicitly or use a main session with a previous channel."
},
"fallbackUsed": false,
"delivered": false
},
"sessionId": "ade3f81b-ab2e-442f-841a-4faafeb687dd",
"sessionKey": "agent:main:main"
},

deleted

woeful treeBOT Apr 29, 2026, 12:23 PM

#

Aha — this one did run successfully:

"status": "ok",
"durationMs": 20616

The problem is delivery, not the agent/cron execution:

"deliveryStatus": "not-delivered",
"error": "Channel is required (no configured channels detected). Set delivery.channel explicitly or use a main session with a previous channel."

So the cron job was configured with:

#

"channel": "last",
"to": null

That’s fragile: it only works if OpenClaw can resolve a previous “last channel” route for that session. Looks like that route got lost/reset sometime between yesterday (April 28) and today (April 29).

Fix it by giving the cron job an explicit delivery target:

openclaw cron edit 2eee5f9c-4664-4c41-948d-1939655a9a87 \
  --announce \
  --channel discord \
  --to "channel:<discord-channel-or-thread-id>"

For this thread specifically, the target id appears to be:

#

--to "channel:1499021752456446104"

If you deleted that job already, same idea when recreating it: don’t use channel: "last" for monitor jobs; use explicit --channel discord --to "channel:...".

Docs: https://docs.openclaw.ai/automation/cron-jobs#delivery-and-output

meager flint Apr 29, 2026, 12:24 PM

#

i'm using exclusively the control ui no channels

woeful treeBOT Apr 29, 2026, 12:25 PM

#

Ah, got it — then the fix is the opposite of what I suggested for Discord delivery.

If you use only Control UI / WebChat, don’t use cron announce delivery. announce is for outbound channel providers like Discord/Telegram/Slack/etc. The Control UI isn’t treated as a configurable outbound channel target there, so channel: "last" resolves to nothing and you get:

Channel is required (no configured channels detected)

For Control UI-only monitor jobs, set delivery to None / no-deliver:

openclaw cron edit <job-id> --no-deliver

Or in Control UI: Cron → Edit job → Delivery → None.

The run results should still be visible in Cron run history / task/session history; it just won’t try to “post” anywhere.

#

If these were older jobs/migrated jobs, also worth running:

openclaw doctor --fix

because older cron delivery fields can get normalized by doctor. Docs: https://docs.openclaw.ai/cli/cron#migrating-older-jobs

meager flint Apr 29, 2026, 12:27 PM

#

also has plugins i didn't enable, it acts like it's typing a response then immediately quits and sends me nothing, i switched from gemma to GPT (5.5) and no change

woeful treeBOT Apr 29, 2026, 12:27 PM

#

That points away from the model. If Gemma and GPT both do it, it’s probably routing/session/plugin/config, not inference.

Two things:

Plugins showing up can be normal. Bundled plugins ship with OpenClaw; some are enabled by default, especially provider/browser/speech-ish ones. To see what’s actually enabled:

openclaw plugins list --enabled --verbose
openclaw plugins doctor

Typing then nothing usually means the run started, then either:

the assistant final was empty / NO_REPLY and got suppressed,
the run errored before final output,
or this is a cron/no-deliver run, so it only appears in run history.

Please grab a live log while reproducing a normal Control UI chat message:

#

openclaw --version
openclaw logs --follow

Then send “say hello” in Control UI and paste the log lines from that moment, especially anything with chat.send, agent, NO_REPLY, error, lifecycle, or delivery.

Also run:

openclaw status --all

I’m especially suspicious of a silent/empty final response or plugin/runtime config weirdness, not the model choice.

meager flint Apr 29, 2026, 12:31 PM

#

https://pastebin.com/mw02rZMZ

woeful treeBOT Apr 29, 2026, 12:32 PM

#

Found it. The actual failure is:

ollama/gemma4:e4b returns an incomplete terminal response.
OpenClaw falls back to openai/gpt-5.5.
OpenAI then fails with quota/rate limit / billing.
Final line: “All models failed (2)”. That’s why it looks like it starts typing and then sends nothing. (pastebin.com)

So switching from Gemma to GPT didn’t help because GPT is currently failing with:

You exceeded your current quota, please check your plan and billing details

Fix path:

#

openclaw models status

Then either:

fix the OpenAI billing/quota/key for that profile, or
remove GPT as fallback and choose a working provider/model, or
fix/replace the local Ollama model, since gemma4:e4b is returning incomplete output.

The plugins are probably not the cause: the log says several plugin entries are disabled but config is present, not that they’re actively running. (pastebin.com)

Also: your old cron failures were delivery config (announce -> last with no channel route), but those jobs appear removed now. Main current blocker is model chain exhausted.

meager flint Apr 29, 2026, 12:44 PM

#

wrong model selected

woeful treeBOT Apr 29, 2026, 12:44 PM

#

Yep — that explains the “typing then nothing” perfectly.

In Control UI, check two places:

Chat header model picker
That writes a persistent per-session override, not just one turn. Pick the correct model there, then send a tiny test like say hello.
Cron job model override
For each monitor job: Cron → Edit job → Agent/model settings and clear or change the model override. Cron can keep using a model different from the chat default.

CLI sanity checks:

openclaw models status
openclaw sessions --verbose
openclaw cron show <job-id>

#

To change the global default:

openclaw models set <provider/model>
openclaw gateway restart

Important nuance: a user-selected/session-selected model is strict — if that selected model fails, OpenClaw won’t silently use some unrelated fallback. That’s why the wrong selected model can make the whole UI look broken.

Docs: https://docs.openclaw.ai/concepts/models

meager flint Apr 29, 2026, 12:48 PM

#

good to close this

#Agent not responding