repo stringclasses 1
value | github_id int64 3.77B 4.38B | github_node_id stringlengths 18 24 | number int64 75 77.8k | html_url stringlengths 46 49 | api_url stringlengths 56 59 | title stringlengths 1 746 | body stringlengths 1 55.7k ⌀ | state stringclasses 2
values | state_reason stringclasses 4
values | locked bool 2
classes | comments_count int64 0 128 | labels listlengths 0 5 | assignees listlengths 0 2 | created_at stringdate 2026-01-01 16:55:34 2026-05-05 11:59:03 | updated_at stringdate 2026-03-24 00:13:29 2026-05-05 11:59:15 | closed_at stringlengths 20 20 ⌀ | author_association stringclasses 3
values | milestone_title stringclasses 0
values | snapshot_id stringclasses 76
values | extracted_at stringdate 2026-03-24 23:36:49 2026-05-05 12:00:33 | author_login stringlengths 3 38 | author_id int64 346 282M | author_node_id stringlengths 12 20 | author_type stringclasses 2
values | author_site_admin bool 2
classes |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
openclaw/openclaw | 4,292,650,034 | I_kwDOQb6kR87_3KQy | 69,151 | https://github.com/openclaw/openclaw/issues/69151 | https://api.github.com/repos/openclaw/openclaw/issues/69151 | [Bug] session_status tool reports 0% context — contextTokens not passed to buildStatusText() | ## Bug Description
The `session_status` tool always reports `0/131k (0%)` context usage, even in sessions with substantial content. The `/status` text command works correctly, but the tool (used by agents via `session_status`) does not.
## Root Cause
In `src/agents/tools/session-status-tool.ts`, the `buildStatusText... | closed | not_planned | true | 1 | [] | [] | 2026-04-20T01:54:19Z | 2026-04-28T04:49:45Z | 2026-04-25T18:48:56Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | upchan-agent | 261,923,048 | U_kgDOD5yg6A | User | false |
openclaw/openclaw | 4,059,364,525 | I_kwDOQb6kR87x9Pyt | 43,351 | https://github.com/openclaw/openclaw/issues/43351 | https://api.github.com/repos/openclaw/openclaw/issues/43351 | [Bug] Cron fails to start: TypeError Cannot read properties of undefined (reading 'kind') | ## Bug Description
Cron functionality fails to start in OpenClaw Gateway (both Dashboard and CLI).
## Environment
- OpenClaw version: 2026.3.8
- OS: Linux (Ubuntu)
- Node: v22.22.0
## Steps to Reproduce
1. Open Gateway Dashboard (http://127.0.0.1:18789/)
2. Go to Cron section
3. Try to create a new cron job
4. Or run... | closed | completed | true | 2 | [] | [] | 2026-03-11T16:44:55Z | 2026-04-28T05:24:37Z | 2026-04-25T10:29:31Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | linsipeng | 20,126,238 | MDQ6VXNlcjIwMTI2MjM4 | User | false |
openclaw/openclaw | 4,292,734,253 | I_kwDOQb6kR87_3e0t | 69,156 | https://github.com/openclaw/openclaw/issues/69156 | https://api.github.com/repos/openclaw/openclaw/issues/69156 | Status false-positive plugin allowlist/provenance warnings in 2026.4.15 from memory-search cold plugin loads | Observed on installed OpenClaw 2026.4.15 on macOS arm64.
Symptom
- `openclaw status`
- `openclaw status --deep`
can emit false warnings like:
- `[plugins] plugins.allow is empty ...`
- `[plugins] lossless-claw: loaded without install/load-path provenance ...`
This happened even when:
- `~/.openclaw/openclaw.json` ha... | closed | completed | true | 1 | [] | [] | 2026-04-20T02:26:10Z | 2026-04-23T04:22:25Z | 2026-04-20T19:31:37Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | RogerGimbel | 204,254,556 | U_kgDODCytXA | User | false |
openclaw/openclaw | 4,292,748,395 | I_kwDOQb6kR87_3iRr | 69,157 | https://github.com/openclaw/openclaw/issues/69157 | https://api.github.com/repos/openclaw/openclaw/issues/69157 | Slack Socket Mode stale-socket health uses inbound app-event recency as transport liveness | ## Summary
Slack Socket Mode health currently treats old inbound app-event activity (`lastEventAt`) as evidence that the websocket is stale. On quiet Slack workspaces this creates false-positive `stale-socket` restarts even when the Slack SDK is not reporting websocket keepalive failures.
This should be treated as a ... | closed | completed | true | 1 | [] | [] | 2026-04-20T02:31:39Z | 2026-04-24T04:25:52Z | 2026-04-22T02:33:20Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | bill492 | 257,063,863 | U_kgDOD1J7tw | User | false |
openclaw/openclaw | 4,292,772,661 | I_kwDOQb6kR87_3oM1 | 69,160 | https://github.com/openclaw/openclaw/issues/69160 | https://api.github.com/repos/openclaw/openclaw/issues/69160 | [Bug]: Onboarding multiple providers overwrites agents.defaults.models | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When onboarding multiple providers sequentially, the second provider's `configPatch` replaces `agents.defaults.models` entirely instead of merging with models from the first provider.
### Steps to reproduce
... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-04-20T02:40:53Z | 2026-04-28T04:49:44Z | 2026-04-25T18:49:09Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rollcarry | 57,675,145 | MDQ6VXNlcjU3Njc1MTQ1 | User | false |
openclaw/openclaw | 4,292,778,267 | I_kwDOQb6kR87_3pkb | 69,161 | https://github.com/openclaw/openclaw/issues/69161 | https://api.github.com/repos/openclaw/openclaw/issues/69161 | Cron-context exec denials should exit fast, not thrash verbose | ## Summary
When an `openclaw cron` fires and the agent tries to `exec` a command that requires interactive approval (ask-fallback=allowlist, not in allowlist), the gateway emits `exec denied: Cron runs cannot wait for interactive exec approval`. The cron's LLM doesn't exit — it generates a 20+ line "Blocked: evidence/... | closed | completed | true | 1 | [] | [] | 2026-04-20T02:42:56Z | 2026-04-29T04:37:32Z | 2026-04-26T10:39:26Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | fawadafr | 1,020,421 | MDQ6VXNlcjEwMjA0MjE= | User | false |
openclaw/openclaw | 4,292,778,445 | I_kwDOQb6kR87_3pnN | 69,162 | https://github.com/openclaw/openclaw/issues/69162 | https://api.github.com/repos/openclaw/openclaw/issues/69162 | openclaw cron run (dry-run) doesn't match scheduled fire semantics under exec approval policy | ## Summary
`openclaw cron run <id>` (manual/dry-run) evaluates exec approvals under a different effective policy than the same cron firing on its schedule. Specifically: commands that pass during `cron run` can be denied during scheduled fire with `exec denied: Cron runs cannot wait for interactive exec approval`.
Th... | closed | completed | true | 1 | [] | [] | 2026-04-20T02:43:00Z | 2026-04-28T04:49:40Z | 2026-04-25T19:04:49Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | fawadafr | 1,020,421 | MDQ6VXNlcjEwMjA0MjE= | User | false |
openclaw/openclaw | 4,145,245,224 | I_kwDOQb6kR873E2wo | 55,259 | https://github.com/openclaw/openclaw/issues/55259 | https://api.github.com/repos/openclaw/openclaw/issues/55259 | Feature Request: Add Gemini TTS Provider (Google AI Studio) | ## Summary
Add Gemini TTS (`gemini-2.5-flash-preview-tts`) as a built-in TTS provider alongside Microsoft Edge TTS, OpenAI, and ElevenLabs.
## Why This Matters
- **Free tier available** - Gemini TTS offers a free tier via Google AI Studio
- **High quality voices** - Competitive with OpenAI/ElevenLabs
- **Already usi... | closed | completed | true | 2 | [] | [] | 2026-03-26T15:56:22Z | 2026-04-28T05:24:26Z | 2026-04-25T10:59:33Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mizoz | 56,456,861 | MDQ6VXNlcjU2NDU2ODYx | User | false |
openclaw/openclaw | 4,292,808,416 | I_kwDOQb6kR87_3w7g | 69,165 | https://github.com/openclaw/openclaw/issues/69165 | https://api.github.com/repos/openclaw/openclaw/issues/69165 | [Feature]: Outbound Telegram send queue with per-chat rate limiting and retry_after-aware backoff | ---
title: "[Feature]: Outbound Telegram send queue with per-chat rate limiting and retry_after-aware backoff"
labels: ["enhancement"]
---
## Summary
Add an outbound message queue for the Telegram channel that enforces per-chat rate limits and respects Telegram's `retry_after` parameter on 429 responses, with exponen... | open | null | false | 2 | [] | [] | 2026-04-20T02:53:41Z | 2026-05-03T22:30:05Z | null | NONE | null | 20260504T000036Z | 2026-05-04T00:00:36Z | r3n3x | 23,291,787 | MDQ6VXNlcjIzMjkxNzg3 | User | false |
openclaw/openclaw | 4,292,824,362 | I_kwDOQb6kR87_300q | 69,166 | https://github.com/openclaw/openclaw/issues/69166 | https://api.github.com/repos/openclaw/openclaw/issues/69166 | sessions_spawn: streamTo schema lacks runtime=acp hint, causes 2x-token retry loop for subagent callers | ## Summary
When a subagent (runtime=`subagent`) calls `sessions_spawn`, the model routinely includes `streamTo: "parent"` in the arguments. The server correctly rejects this with:
```
streamTo is only supported for runtime=acp; got runtime=subagent
```
The model then retries without `streamTo` and the second call su... | closed | not_planned | true | 1 | [] | [] | 2026-04-20T02:59:30Z | 2026-04-29T04:37:31Z | 2026-04-26T10:39:31Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | IsaacAPerez | 54,831,668 | MDQ6VXNlcjU0ODMxNjY4 | User | false |
openclaw/openclaw | 4,292,829,049 | I_kwDOQb6kR87_3195 | 69,167 | https://github.com/openclaw/openclaw/issues/69167 | https://api.github.com/repos/openclaw/openclaw/issues/69167 | [Feature]: Automated bug-fix pipeline powered by Claude Code for community contributions | ### Summary
Build an automated / semi-automated bug-fix pipeline using Claude Code (AI agent) to help the community triage, reproduce, fix, and submit PRs for the rapidly growing backlog of open bugs.
### Problem to solve
The issue backlog is growing faster than the community can address:
- **2,727 open bugs** as o... | closed | not_planned | true | 1 | [] | [] | 2026-04-20T03:00:59Z | 2026-04-28T04:49:38Z | 2026-04-25T19:04:55Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | skylee-01 | 46,676,799 | MDQ6VXNlcjQ2Njc2Nzk5 | User | false |
openclaw/openclaw | 4,292,840,857 | I_kwDOQb6kR87_342Z | 69,171 | https://github.com/openclaw/openclaw/issues/69171 | https://api.github.com/repos/openclaw/openclaw/issues/69171 | [Bug] media-understanding ignores agents.defaults.imageModel, routes images to agent model instead | ## Bug type
Regression — `agents.defaults.imageModel` config is effectively dead code for the `media-understanding` subsystem.
## Summary
When a text-only model is configured as the agent model (e.g., `zai/glm-5`) and a separate vision model is configured via `agents.defaults.imageModel` (e.g., `zai/glm-4.6v`), the `... | closed | completed | true | 4 | [] | [] | 2026-04-20T03:04:43Z | 2026-04-26T04:29:19Z | 2026-04-24T01:23:43Z | NONE | null | 20260426T060034Z | 2026-04-26T06:00:34Z | xcheljd | 203,027,001 | U_kgDODBnyOQ | User | false |
openclaw/openclaw | 4,292,979,049 | I_kwDOQb6kR87_4alp | 69,180 | https://github.com/openclaw/openclaw/issues/69180 | https://api.github.com/repos/openclaw/openclaw/issues/69180 | [Bug]: Unknown memory embedding provider: github-copilot | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Unable to use github-copilot as memory embedding provider.
OpenClaw 2026.4.15 (041266a)
### Steps to reproduce
1. Add Config to openclaw.json
```
{
"agents": {
"defaults": {
"memorySearch": {
... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-04-20T03:54:03Z | 2026-04-28T04:49:37Z | 2026-04-25T19:05:02Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mannbadal | 108,790,470 | U_kgDOBnwCxg | User | false |
openclaw/openclaw | 4,292,980,326 | I_kwDOQb6kR87_4a5m | 69,181 | https://github.com/openclaw/openclaw/issues/69181 | https://api.github.com/repos/openclaw/openclaw/issues/69181 | plugins install incorrectly treats npm packages as hook packs | ## Bug Description
`openclaw plugins install @martian-engineering/lossless-claw` fails with error:
```
Also not a valid hook pack: Error: package.json missing openclaw.hooks
```
The package has valid `openclaw.extensions` in package.json, but the CLI appears to incorrectly attempt hook pack validation first, then fai... | closed | duplicate | true | 1 | [
"duplicate",
"dedupe:child"
] | [] | 2026-04-20T03:54:24Z | 2026-04-28T04:36:57Z | 2026-04-26T03:50:38Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ygc3817922006-sketch | 266,983,010 | U_kgDOD-nWYg | User | false |
openclaw/openclaw | 4,292,984,657 | I_kwDOQb6kR87_4b9R | 69,182 | https://github.com/openclaw/openclaw/issues/69182 | https://api.github.com/repos/openclaw/openclaw/issues/69182 | [Performance] active-memory plugin can block chat replies up to ~5s | ## Summary
Chat replies were laggy (~5 seconds before response).
Investigation showed the `active-memory` plugin could block the main response until its timeout.
## What I observed
- With `active-memory` enabled, a simple message returned after ~5s.
- After disabling `active-memory`, responses became noticeably fast i... | open | null | false | 1 | [] | [] | 2026-04-20T03:55:51Z | 2026-05-02T16:06:49Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | Kru442 | 187,809,917 | U_kgDOCzHAfQ | User | false |
openclaw/openclaw | 4,292,997,257 | I_kwDOQb6kR87_4fCJ | 69,183 | https://github.com/openclaw/openclaw/issues/69183 | https://api.github.com/repos/openclaw/openclaw/issues/69183 | [Bug]: Local outbound media allowlist/staging regression breaks attachments and screenshots across channels | ### Bug type
Regression / behavior bug
### Beta release blocker
No
### Summary
There appears to be a shared regression / design gap in the **local outbound media pipeline** affecting multiple surfaces:
- Telegram final assistant replies using `MEDIA:`
- WhatsApp agent replies with images/files
- browser / screens... | closed | completed | true | 2 | [] | [] | 2026-04-20T03:59:38Z | 2026-04-30T04:38:56Z | 2026-04-26T05:25:10Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | edwincrug | 6,675,460 | MDQ6VXNlcjY2NzU0NjA= | User | false |
openclaw/openclaw | 4,293,012,285 | I_kwDOQb6kR87_4is9 | 69,184 | https://github.com/openclaw/openclaw/issues/69184 | https://api.github.com/repos/openclaw/openclaw/issues/69184 | [Feature Request] feishu_drive tool: add download action for cloud storage files | ## Is your feature request related to a problem?
The `feishu_drive` tool currently supports `list`, `info`, `create_folder`, `move`, and `delete` actions, but has no ability to **download files** from Feishu cloud storage (PDF, .docx, .xlsx attachments uploaded to folders).
This means bots can list and discover files... | closed | not_planned | true | 1 | [] | [] | 2026-04-20T04:04:10Z | 2026-04-30T04:38:56Z | 2026-04-26T05:25:15Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | hollowhn | 202,040,661 | U_kgDODArlVQ | User | false |
openclaw/openclaw | 4,293,018,570 | I_kwDOQb6kR87_4kPK | 69,185 | https://github.com/openclaw/openclaw/issues/69185 | https://api.github.com/repos/openclaw/openclaw/issues/69185 | Control UI: allow non-image file uploads (documents, zips, archives) | **Version**: `openclaw@2026.4.15`
**Surface**: Control UI (webchat)
## Current behavior
The Control UI file picker restricts attachments to image MIME types only. Inspecting the bundled UI (`dist/control-ui/assets/index-*.js`) shows the accept list is limited to:
```
image/*, image/gif, image/heic, image/heif, image/j... | closed | not_planned | true | 1 | [] | [] | 2026-04-20T04:05:48Z | 2026-04-29T04:37:30Z | 2026-04-26T10:39:36Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | d-rubik | 57,096,034 | MDQ6VXNlcjU3MDk2MDM0 | User | false |
openclaw/openclaw | 4,293,025,198 | I_kwDOQb6kR87_4l2u | 69,186 | https://github.com/openclaw/openclaw/issues/69186 | https://api.github.com/repos/openclaw/openclaw/issues/69186 | Feature Request: Add completion/success notification sound for agent turns | ## Feature Request: Add completion notification sound
### Problem / User Need
When an agent completes a turn (finishes generating a response), users have no audible indication that the response is ready. Hermes Agent's WebUI provides a notification sound when tasks complete, which is very useful for users who keep the... | open | null | false | 1 | [] | [] | 2026-04-20T04:07:29Z | 2026-05-02T17:31:33Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | canmaxice-maker | 274,350,325 | U_kgDOEFpA9Q | User | false |
openclaw/openclaw | 4,292,706,157 | I_kwDOQb6kR87_3X9t | 69,155 | https://github.com/openclaw/openclaw/issues/69155 | https://api.github.com/repos/openclaw/openclaw/issues/69155 | Session reset keeps old transcript file, causing context bloat across days | I've been running OpenClaw in production with `session.reset.mode: "daily"` for a while, and I noticed something that I think might be intentional design, but the current implementation is incomplete.
## What I observed
After the daily reset fires at 7am, the new session loads the **entire transcript from the old ses... | closed | completed | true | 1 | [] | [] | 2026-04-20T02:15:54Z | 2026-04-28T04:49:45Z | 2026-04-25T18:49:03Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | qq816aa | 265,676,137 | U_kgDOD9XlaQ | User | false |
openclaw/openclaw | 4,293,065,762 | I_kwDOQb6kR87_4vwi | 69,187 | https://github.com/openclaw/openclaw/issues/69187 | https://api.github.com/repos/openclaw/openclaw/issues/69187 | memory-core dreaming plugin: deleteSession() fails 100% → sessions.json leak → gateway OOM | # Dreaming plugin: `deleteSession()` fails on every cron invocation, leaking session entries
## Summary
The `memory-core` dreaming narrative plugin (`dreaming-narrative-BFx7ugb_.js`) runs a 3-phase cron job (light/REM/deep) that **100% leaks its subagent session entries** because `deleteSession()` in the `finally` bl... | closed | not_planned | true | 1 | [] | [] | 2026-04-20T04:22:07Z | 2026-04-29T04:36:29Z | 2026-04-26T12:05:25Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | r0gerliu | 70,790,450 | MDQ6VXNlcjcwNzkwNDUw | User | false |
openclaw/openclaw | 4,293,069,826 | I_kwDOQb6kR87_4wwC | 69,188 | https://github.com/openclaw/openclaw/issues/69188 | https://api.github.com/repos/openclaw/openclaw/issues/69188 | Cron Alert: Gemini CLI bridge OAuth pool | Pool: 4/4 available (100%), 0 in cooldown, 0 busy | closed | not_planned | true | 1 | [] | [] | 2026-04-20T04:23:37Z | 2026-04-28T04:49:37Z | 2026-04-25T19:05:08Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | skdung | 262,492,811 | U_kgDOD6VSiw | User | false |
openclaw/openclaw | 4,293,071,500 | I_kwDOQb6kR87_4xKM | 69,189 | https://github.com/openclaw/openclaw/issues/69189 | https://api.github.com/repos/openclaw/openclaw/issues/69189 | Cron subagent-dispatcher: opus47 reasoning leaks into delivered text; no clean post-processing path without losing channel card rendering | # Cron subagent-dispatcher pattern: opus47 preamble leaks into delivery, no clean escape hatch
## Summary
When an isolated cron job uses a dispatcher-pattern (parent agent spawns a subagent for the real work and relies on `SUBAGENT_FOLLOWUP_HINTS` for delivery replacement), **Claude Opus 4.7's reasoning leaks into th... | closed | not_planned | true | 1 | [] | [] | 2026-04-20T04:24:12Z | 2026-04-22T04:30:26Z | 2026-04-20T04:25:17Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | SuperJony | 15,604,104 | MDQ6VXNlcjE1NjA0MTA0 | User | false |
openclaw/openclaw | 4,293,076,254 | I_kwDOQb6kR87_4yUe | 69,190 | https://github.com/openclaw/openclaw/issues/69190 | https://api.github.com/repos/openclaw/openclaw/issues/69190 | Frequent Timeouts & Agent Ignoring Tool Restrictions (Browser vs. Fetch/Curl) | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
The agent frequently times out when using the `browser` tool and actively bypasses strict `.md` instructions by autonomously using `curl` after the `fetch` tool is disabled.
### Steps to reproduce
1. Start the agent using ... | open | null | false | 1 | [
"bug"
] | [] | 2026-04-20T04:26:01Z | 2026-05-02T16:01:48Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | LEELEE412 | 78,653,980 | MDQ6VXNlcjc4NjUzOTgw | User | false |
openclaw/openclaw | 3,998,341,748 | I_kwDOQb6kR87uUdp0 | 28,037 | https://github.com/openclaw/openclaw/issues/28037 | https://api.github.com/repos/openclaw/openclaw/issues/28037 | Slack Socket Mode: Messages not routing to agent despite Socket Mode connection | ## Bug Description
Slack Socket Mode connects successfully and confirms connection in gateway logs, but messages sent to the bot (both DMs and @mentions in channels) are not being received/routed to the agent.
## Environment
- **OpenClaw version:** 2026.2.25
- **Platform:** macOS (Intel iMac)
- **Node version:** 22.... | closed | completed | true | 17 | [
"stale"
] | [] | 2026-02-26T23:56:36Z | 2026-04-27T05:03:58Z | 2026-04-24T19:08:40Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Dutch-Forward | 264,284,153 | U_kgDOD8Cn-Q | User | false |
openclaw/openclaw | 4,009,419,201 | I_kwDOQb6kR87u-uHB | 31,331 | https://github.com/openclaw/openclaw/issues/31331 | https://api.github.com/repos/openclaw/openclaw/issues/31331 | [Bug]: Docker Install + Sandbox can't workspaceAccess at all | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When running the Gateway inside a Docker container and using Docker-outside-of-Docker for Sandboxing, /workspace can't binds properly.
The Gateway containder attempts to mount the agent workspace using its **internal** path (e.g., /home/no... | open | null | false | 8 | [
"bug",
"dedupe:parent",
"bug:behavior"
] | [] | 2026-03-02T05:47:25Z | 2026-05-05T02:56:36Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | jiesou | 84,175,239 | MDQ6VXNlcjg0MTc1MjM5 | User | false |
openclaw/openclaw | 4,017,925,357 | I_kwDOQb6kR87vfKzt | 33,329 | https://github.com/openclaw/openclaw/issues/33329 | https://api.github.com/repos/openclaw/openclaw/issues/33329 | Document and add config toggles for all implicit discovery mechanisms | ## Related
References #33327 (implicit provider auto-discovery bug)
## Summary
Four implicit discovery mechanisms make network calls on gateway startup based on env vars: Bedrock (`AWS_PROFILE`), Copilot (`GH_TOKEN`), HuggingFace (`HF_TOKEN`), and Ollama. When these calls fail — expired tokens, unreachable servers, ... | open | null | false | 5 | [] | [] | 2026-03-03T17:40:54Z | 2026-05-05T02:56:51Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | bhalliburton | 94,708 | MDQ6VXNlcjk0NzA4 | User | false |
openclaw/openclaw | 4,020,013,372 | I_kwDOQb6kR87vnIk8 | 33,945 | https://github.com/openclaw/openclaw/issues/33945 | https://api.github.com/repos/openclaw/openclaw/issues/33945 | Feature request: config to disable implicit thread mention in Slack channels | ## Problem
In multi-agent (multi-lobster) deployments on Slack, once an agent posts in a thread, **all subsequent messages** in that thread trigger the agent — even without an explicit `@mention`. This is because of the implicit thread mention mechanism (`hasSlackThreadParticipation`).
This causes agents to respond t... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T04:49:13Z | 2026-04-22T04:30:18Z | 2026-04-20T04:27:01Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | Arxchibobo | 212,881,651 | U_kgDODLBQ8w | User | false |
openclaw/openclaw | 4,020,013,780 | I_kwDOQb6kR87vnIrU | 33,946 | https://github.com/openclaw/openclaw/issues/33946 | https://api.github.com/repos/openclaw/openclaw/issues/33946 | Mattermost image context gap (runtime) + message send buffer mismatch | ## Summary
From real-world usage (OpenClaw `2026.2.26` + Mattermost plugin), two image-related problems were observed:
1. **Inbound Mattermost image not reaching effective model image context** (even though file is downloaded).
2. **`message action=send` does not honor `buffer` attachment payloads**.
After reading c... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T04:49:25Z | 2026-04-22T04:30:18Z | 2026-04-20T04:27:02Z | CONTRIBUTOR | null | 20260422T060039Z | 2026-04-22T06:00:39Z | LyleLiu666 | 31,182,860 | MDQ6VXNlcjMxMTgyODYw | User | false |
openclaw/openclaw | 4,020,014,599 | I_kwDOQb6kR87vnI4H | 33,947 | https://github.com/openclaw/openclaw/issues/33947 | https://api.github.com/repos/openclaw/openclaw/issues/33947 | Gateway full process restart fails with spawnSync launchctl ETIMEDOUT, leaving LaunchAgent unloaded | ## Summary
When a config change requires a full gateway restart (e.g. `auth.profiles` modification), the restart mechanism via `spawnSync launchctl` frequently times out (~30s), leaving the LaunchAgent in an unloaded state. Since the agent is no longer registered with launchd, `KeepAlive: true` cannot function, and th... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T04:49:49Z | 2026-04-22T04:30:17Z | 2026-04-20T04:27:04Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | TubeLiu | 14,797,490 | MDQ6VXNlcjE0Nzk3NDkw | User | false |
openclaw/openclaw | 4,020,033,239 | I_kwDOQb6kR87vnNbX | 33,948 | https://github.com/openclaw/openclaw/issues/33948 | https://api.github.com/repos/openclaw/openclaw/issues/33948 | Add message reaction support to feishu channel (add/list reactions) | The feishu extension already has `addReactionFeishu` and `listReactionsFeishu` functions implemented in `src/reactions.ts`, but they are not exposed as tools that agents can use.
**Requested:**
- Expose `addReactionFeishu` as a message tool action
- Expose `listReactionsFeishu` as a message tool action
**Use case:**
... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T04:56:15Z | 2026-04-22T04:30:16Z | 2026-04-20T04:27:06Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | 1205567471-rgb | 264,598,806 | U_kgDOD8V1Fg | User | false |
openclaw/openclaw | 4,020,035,201 | I_kwDOQb6kR87vnN6B | 33,949 | https://github.com/openclaw/openclaw/issues/33949 | https://api.github.com/repos/openclaw/openclaw/issues/33949 | Model Switch Reports Success But Probe Shows 'Running But Not Reachable' | # Bug Report: Model Switch Reports Success But Probe Shows "Running But Not Reachable"
**Date:** 2026-03-03
**OpenClaw Version:** 2026.3.2 (85377a2)
**OS:** macOS 26.3 (arm64)
**Node:** v22.22.0
## Summary
When switching models via `session_status` (e.g., switching to `llama3.1`), the operation reports success, but ... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-04T04:56:56Z | 2026-04-22T04:30:15Z | 2026-04-20T04:27:08Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | joemcurry | 1,192,832 | MDQ6VXNlcjExOTI4MzI= | User | false |
openclaw/openclaw | 4,020,045,454 | I_kwDOQb6kR87vnQaO | 33,951 | https://github.com/openclaw/openclaw/issues/33951 | https://api.github.com/repos/openclaw/openclaw/issues/33951 | [Bug]: Slack typing/status bubble shown for messages that end NO_REPLY (regression in 2026.3.2) | ### Bug type
Regression (worked before, now fails)
### Summary
In v2026.3.2, Slack channels with requireMention: false show an animated typing/status
bubble for every incoming message — even when the agent ultimately produces NO_REPLY
(silent gate). This did not occur in v2026.3.1.
### Steps to reproduce
1.... | closed | not_planned | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T05:00:08Z | 2026-04-22T04:30:14Z | 2026-04-20T04:27:10Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | teosborne | 4,041,297 | MDQ6VXNlcjQwNDEyOTc= | User | false |
openclaw/openclaw | 4,020,076,839 | I_kwDOQb6kR87vnYEn | 33,962 | https://github.com/openclaw/openclaw/issues/33962 | https://api.github.com/repos/openclaw/openclaw/issues/33962 | [Feature]: slug-generator: use lightweight model instead of agent primary to prevent lane congestion | ## Problem
The LLM slug generator (`llm-slug-generator.ts`) generates session memory filenames by calling `resolveAgentEffectiveModelPrimary()` — meaning it always uses the **default agent's primary model** for a trivial 2-word filename task.
This causes two serious issues:
### 1. Heavy model for trivial task
Using ... | open | null | false | 4 | [] | [] | 2026-03-04T05:09:46Z | 2026-05-05T03:11:05Z | null | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | zhangsensen | 26,587,484 | MDQ6VXNlcjI2NTg3NDg0 | User | false |
openclaw/openclaw | 4,020,106,112 | I_kwDOQb6kR87vnfOA | 33,974 | https://github.com/openclaw/openclaw/issues/33974 | https://api.github.com/repos/openclaw/openclaw/issues/33974 | [Bug]: WhatsApp group config changes (requireMention, group policy) silently ignored until channel restart — on-message handler holds stale config reference | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
### Description
`createWebOnMessageHandler` in `src/web/auto-reply/monitor/on-message.ts` captures `params.cfg` at handler creation time. All downstream calls — `applyGroupGating`, `processMessage`, `updateLastRouteInBackground`, `maybeBro... | closed | not_planned | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-04T05:18:31Z | 2026-04-22T04:30:14Z | 2026-04-20T04:27:13Z | CONTRIBUTOR | null | 20260422T060039Z | 2026-04-22T06:00:39Z | asyncjason | 4,665,758 | MDQ6VXNlcjQ2NjU3NTg= | User | false |
openclaw/openclaw | 4,020,116,757 | I_kwDOQb6kR87vnh0V | 33,978 | https://github.com/openclaw/openclaw/issues/33978 | https://api.github.com/repos/openclaw/openclaw/issues/33978 | [Bug] Discord: Cannot send media/files to user DM - Unknown Channel error | ## Bug Description
When trying to send a file attachment to a Discord user via DM using `openclaw message send --channel discord --account <account> --target <userId> --media <file>`, it returns error:
```
Unknown Channel
```
## Expected Behavior
- Sending files to Discord user IDs should work (like Telegram does)
-... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-04T05:21:42Z | 2026-04-22T04:30:13Z | 2026-04-20T04:27:15Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | locolyric | 320,045 | MDQ6VXNlcjMyMDA0NQ== | User | false |
openclaw/openclaw | 4,020,219,367 | I_kwDOQb6kR87vn63n | 34,006 | https://github.com/openclaw/openclaw/issues/34006 | https://api.github.com/repos/openclaw/openclaw/issues/34006 | Gateway crashes on unhandled fetch rejection when node disconnects | ## Description
The gateway crashes with an unhandled promise rejection when a node disconnects (e.g., during reboot).
## Error
\\\
[clawdbot] Unhandled promise rejection: TypeError: fetch failed
at node:internal/deps/undici/undici:15482:13
at processTicksAndRejections (node:internal/process/task_queues:105:5)
... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T05:47:38Z | 2026-04-22T04:30:12Z | 2026-04-20T04:27:19Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | foxsake123 | 70,024,056 | MDQ6VXNlcjcwMDI0MDU2 | User | false |
openclaw/openclaw | 4,020,226,920 | I_kwDOQb6kR87vn8to | 34,008 | https://github.com/openclaw/openclaw/issues/34008 | https://api.github.com/repos/openclaw/openclaw/issues/34008 | [Bug] Gemini 3 function calling softlocked via Ollama Cloud (missing thought_signature) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
[Bug] Gemini 3 function calling softlocked via Ollama Cloud (missing thought_signature)
### Steps to reproduce
1. Configure OpenClaw to use Ollama Cloud as the provider.
2. Set model to gemini-3-flash-preview:cloud.
3. Enable tools/functi... | closed | not_planned | true | 6 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-04T05:49:53Z | 2026-04-22T04:30:11Z | 2026-04-20T04:27:21Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | reflexmrl | 221,397,360 | U_kgDODTJBcA | User | false |
openclaw/openclaw | 4,020,282,126 | I_kwDOQb6kR87voKMO | 34,021 | https://github.com/openclaw/openclaw/issues/34021 | https://api.github.com/repos/openclaw/openclaw/issues/34021 | [Bug]: 【UI】【文档网站】使用左上角的语言切换时会默认回退到默认位置，应保持用户实际浏览位置 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
无
### Steps to reproduce
1、浏览器访问：https://docs.openclaw.ai/install/docker
2、将网站语言切换为简体中文
### Expected behavior
保持用户当前浏览位置
### Actual behavior
回退到默认位置
### OpenClaw version
无
### Operating system
Windows11
### Install method
_No re... | closed | not_planned | true | 3 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-04T06:05:09Z | 2026-04-22T04:30:11Z | 2026-04-20T04:27:23Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | ricsy | 80,663,724 | MDQ6VXNlcjgwNjYzNzI0 | User | false |
openclaw/openclaw | 4,020,320,734 | I_kwDOQb6kR87voTne | 34,047 | https://github.com/openclaw/openclaw/issues/34047 | https://api.github.com/repos/openclaw/openclaw/issues/34047 | Webchat: Add confirmation dialog for 'New session' button | It's easy to accidentally trigger the **New session** button via keyboard navigation (Tab + Enter while typing in the chat input). A confirmation prompt (e.g. _"Start new session? Current context will be lost"_) would prevent accidental session resets.
**Steps to reproduce:**
1. Be typing in the webchat input
2. Press... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T06:17:09Z | 2026-04-22T04:30:10Z | 2026-04-20T04:27:26Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | leobon | 26,121,216 | MDQ6VXNlcjI2MTIxMjE2 | User | false |
openclaw/openclaw | 4,020,357,351 | I_kwDOQb6kR87vocjn | 34,054 | https://github.com/openclaw/openclaw/issues/34054 | https://api.github.com/repos/openclaw/openclaw/issues/34054 | ACP gateway should always call runtime.close() for completed oneshot sessions | ## Problem
When an ACP session with `mode: "run"` (oneshot) completes — whether successfully, on error, or on timeout — the gateway does not always call `runtime.close()` on the acpx backend. This leaves the backend session in an "open" state, counting against `maxConcurrentSessions` even though no process is running.... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T06:29:16Z | 2026-04-22T04:30:09Z | 2026-04-20T04:27:28Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | allanjeng | 14,968,942 | MDQ6VXNlcjE0OTY4OTQy | User | false |
openclaw/openclaw | 4,020,362,040 | I_kwDOQb6kR87vods4 | 34,056 | https://github.com/openclaw/openclaw/issues/34056 | https://api.github.com/repos/openclaw/openclaw/issues/34056 | Feature: Send authorization hint when bot is @mentioned but unauthorized | ## Problem
When a bot is added to a group but not authorized (e.g., `groupPolicy: "allowlist"` but no `allowFrom` configured), messages are silently dropped with no feedback to users. Even when the bot is @mentioned, it gives zero response, making it impossible to distinguish between "bot didn't receive the message" a... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T06:31:00Z | 2026-04-22T04:30:08Z | 2026-04-20T04:27:30Z | CONTRIBUTOR | null | 20260422T060039Z | 2026-04-22T06:00:39Z | dongzhenye | 5,765,843 | MDQ6VXNlcjU3NjU4NDM= | User | false |
openclaw/openclaw | 4,020,363,915 | I_kwDOQb6kR87voeKL | 34,057 | https://github.com/openclaw/openclaw/issues/34057 | https://api.github.com/repos/openclaw/openclaw/issues/34057 | Feature: Bot should self-report group ID when added to group | ## Problem
Currently, getting a group's Telegram ID requires external tools (e.g., @userinfobot) or manual operations. This adds friction when configuring new groups in OpenClaw.
The bot already receives the group ID in every message (`msg.chat.id`), so it has the information needed to self-report.
## Proposed Solut... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T06:31:43Z | 2026-04-22T04:30:07Z | 2026-04-20T04:27:32Z | CONTRIBUTOR | null | 20260422T060039Z | 2026-04-22T06:00:39Z | dongzhenye | 5,765,843 | MDQ6VXNlcjU3NjU4NDM= | User | false |
openclaw/openclaw | 4,020,387,208 | I_kwDOQb6kR87voj2I | 34,061 | https://github.com/openclaw/openclaw/issues/34061 | https://api.github.com/repos/openclaw/openclaw/issues/34061 | Feature: Support hot-reload for group configuration changes | ## Problem
When adding a new group to the configuration, users must run `openclaw gateway restart`, which:
- Interrupts all active sessions
- Drops in-flight messages
- Causes service downtime for all channels
This is disruptive, especially when managing multiple groups or testing configurations.
## Current Behavior... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T06:40:10Z | 2026-04-22T04:30:07Z | 2026-04-20T04:27:34Z | CONTRIBUTOR | null | 20260422T060039Z | 2026-04-22T06:00:39Z | dongzhenye | 5,765,843 | MDQ6VXNlcjU3NjU4NDM= | User | false |
openclaw/openclaw | 4,020,387,685 | I_kwDOQb6kR87voj9l | 34,063 | https://github.com/openclaw/openclaw/issues/34063 | https://api.github.com/repos/openclaw/openclaw/issues/34063 | Enhancement: Doctor should verify actual Privacy Mode status via API | ## Problem
`openclaw doctor` warns about Privacy Mode even when it's already disabled in BotFather:
```
Config allows unmentioned group messages (requireMention=false).
Telegram Bot API privacy mode will block most group messages unless disabled.
```
This warning is based on **config inference**, not actual API veri... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T06:40:21Z | 2026-04-22T04:30:06Z | 2026-04-20T04:27:36Z | CONTRIBUTOR | null | 20260422T060039Z | 2026-04-22T06:00:39Z | dongzhenye | 5,765,843 | MDQ6VXNlcjU3NjU4NDM= | User | false |
openclaw/openclaw | 4,020,453,640 | I_kwDOQb6kR87vo0EI | 34,089 | https://github.com/openclaw/openclaw/issues/34089 | https://api.github.com/repos/openclaw/openclaw/issues/34089 | macOS LaunchAgent should set ProcessType=Background to prevent termination during session transitions | ## Bug Description
On macOS, the OpenClaw gateway service gets terminated (SIGTERM) when the user session transitions to the login screen, even when sleep is disabled. The LaunchAgent generated by `openclaw gateway install` is missing the `ProcessType=Background` key, causing macOS to kill the process for "efficiency"... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T07:01:32Z | 2026-04-22T04:30:05Z | 2026-04-20T04:27:38Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | joemcurry | 1,192,832 | MDQ6VXNlcjExOTI4MzI= | User | false |
openclaw/openclaw | 4,020,461,131 | I_kwDOQb6kR87vo15L | 34,092 | https://github.com/openclaw/openclaw/issues/34092 | https://api.github.com/repos/openclaw/openclaw/issues/34092 | [Bug]: Chat "copy button" not working on Windows | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
OpenClaw version: 2026.3.2
The "copy button" of the chat box is not working on Windows with Chrome browser. It should expect to copy the output markdown message.
<img width="773" height="551" alt="Image" src="https://github.com/user-attac... | closed | not_planned | true | 5 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-04T07:04:18Z | 2026-04-22T04:29:17Z | 2026-04-20T04:27:39Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | sweihub | 5,334,841 | MDQ6VXNlcjUzMzQ4NDE= | User | false |
openclaw/openclaw | 4,020,462,499 | I_kwDOQb6kR87vo2Oj | 34,093 | https://github.com/openclaw/openclaw/issues/34093 | https://api.github.com/repos/openclaw/openclaw/issues/34093 | [Bug]: Feishu channel silently drops replies when blockStreamingDefault is on | ## Summary
When `agents.defaults.blockStreamingDefault` is set to `"on"`, Feishu channel replies are silently dropped. The agent generates replies, but they are never delivered to the Feishu chat. Gateway logs show `dispatch complete (queuedFinal=false, replies=0)` for all non-command messages.
## Steps to Reproduce
... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T07:04:47Z | 2026-04-22T04:30:05Z | 2026-04-20T04:27:41Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | TeigenZhang | 38,639,187 | MDQ6VXNlcjM4NjM5MTg3 | User | false |
openclaw/openclaw | 4,020,482,502 | I_kwDOQb6kR87vo7HG | 34,097 | https://github.com/openclaw/openclaw/issues/34097 | https://api.github.com/repos/openclaw/openclaw/issues/34097 | [Feature Request] Native MCP (Model Context Protocol) server integration | ## Problem Description
OpenClaw currently lacks native support for **Model Context Protocol (MCP)** servers, which are becoming the standard way to expose tools to AI agents (adopted by Claude Desktop, Cursor, Windsurf, etc.).
Users who have configured MCP servers via mcporter CLI expect these tools to be automatical... | closed | not_planned | true | 8 | [
"stale"
] | [] | 2026-03-04T07:11:25Z | 2026-04-27T05:03:21Z | 2026-04-24T19:11:50Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | DuckRaiser | 52,232,447 | MDQ6VXNlcjUyMjMyNDQ3 | User | false |
openclaw/openclaw | 4,020,523,506 | I_kwDOQb6kR87vpFHy | 34,108 | https://github.com/openclaw/openclaw/issues/34108 | https://api.github.com/repos/openclaw/openclaw/issues/34108 | [Bug]: Feishu streaming card renders duplicate text blocks | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When using Feishu card streaming mode (`renderMode: "card"` with `streaming: true`), the same content block is rendered 3+ times in a single message card during streaming updates. The card shows duplicated text that grows with each streamin... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-04T07:23:40Z | 2026-04-27T05:03:20Z | 2026-04-24T19:21:45Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Mickey0811 | 49,522,921 | MDQ6VXNlcjQ5NTIyOTIx | User | false |
openclaw/openclaw | 4,020,559,204 | I_kwDOQb6kR87vpN1k | 34,117 | https://github.com/openclaw/openclaw/issues/34117 | https://api.github.com/repos/openclaw/openclaw/issues/34117 | [Bug]: Anthropic OAuth refresh token discarded by configure; auto-refresh always fails, requires manual re-auth every ~8h | ## Environment
- **Version:** 2026.3.2
- **OS:** macOS
- **Provider:** Anthropic (claude-sonnet-4-6)
## Describe the bug
When running `openclaw configure` for Anthropic, the OAuth flow correctly obtains both an access token and refresh token. However the refresh token is discarded and the auto-renew mechanism never w... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-04T07:34:15Z | 2026-04-22T04:30:03Z | 2026-04-20T04:27:48Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | wiziswiz | 83,570,252 | MDQ6VXNlcjgzNTcwMjUy | User | false |
openclaw/openclaw | 4,020,605,522 | I_kwDOQb6kR87vpZJS | 34,134 | https://github.com/openclaw/openclaw/issues/34134 | https://api.github.com/repos/openclaw/openclaw/issues/34134 | [Bug]: Switching to openai-completions is invalid | ### Bug type
Regression (worked before, now fails)
### Summary
I modified the custom URL and the openai-completions format in the configuration file, but found that openclaw sends requests to the outside in the Claude format instead of the normal OpenAI format, resulting in empty robot responses
```
"models": {
... | closed | not_planned | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T07:48:25Z | 2026-04-22T04:30:02Z | 2026-04-20T04:27:51Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | chung1912 | 38,153,111 | MDQ6VXNlcjM4MTUzMTEx | User | false |
openclaw/openclaw | 4,020,633,626 | I_kwDOQb6kR87vpgAa | 34,144 | https://github.com/openclaw/openclaw/issues/34144 | https://api.github.com/repos/openclaw/openclaw/issues/34144 | [Feature] Plugin hooks to modify tool results before LLM sees them | ## Problem
Plugin hooks currently cannot change the tool result that the LLM sees on the same turn.
- `after_tool_call` — can only read the tool result, not modify it (fire-and-forget)
- `tool_result_persist` — modifies the persisted transcript only; the LLM already received the original result in the current turn
T... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T07:56:51Z | 2026-04-22T04:30:01Z | 2026-04-20T04:27:53Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | juppytt | 46,583,939 | MDQ6VXNlcjQ2NTgzOTM5 | User | false |
openclaw/openclaw | 4,020,673,999 | I_kwDOQb6kR87vpp3P | 34,153 | https://github.com/openclaw/openclaw/issues/34153 | https://api.github.com/repos/openclaw/openclaw/issues/34153 | [Bug]: Sender (untrusted metadata) | ### Bug type
Regression (worked before, now fails)
### Summary
为啥请求LLM的时候会带有：Sender (untrusted metadata):\n```json\n{\n \"label\": \"openclaw-control-ui\",\n \"id\": \"openclaw-control-ui\"\n}\n```\n\n[Wed 2026-03-04 15:56 GMT+8] 这些信息。
### Steps to reproduce
<img width="1619" height="678" alt="Image" src="https:... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T08:07:54Z | 2026-04-22T04:30:01Z | 2026-04-20T04:27:55Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | Dreaming-world | 25,942,191 | MDQ6VXNlcjI1OTQyMTkx | User | false |
openclaw/openclaw | 4,020,678,523 | I_kwDOQb6kR87vpq97 | 34,155 | https://github.com/openclaw/openclaw/issues/34155 | https://api.github.com/repos/openclaw/openclaw/issues/34155 | WhatsApp provider: stale-socket every ~35 minutes on idle connections (keepalive regression) | ## Summary
WhatsApp connections consistently go stale after ~35 minutes of no message traffic, triggering a `health-monitor: restarting (reason: stale-socket)` cycle. This happens with clockwork precision regardless of network conditions.
## Environment
- **OpenClaw version:** 2026.3.2
- **OS:** Arch Linux, kernel 6... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-04T08:09:03Z | 2026-04-22T04:30:00Z | 2026-04-20T04:27:57Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | haydenmckay | 13,712,068 | MDQ6VXNlcjEzNzEyMDY4 | User | false |
openclaw/openclaw | 4,020,687,324 | I_kwDOQb6kR87vptHc | 34,156 | https://github.com/openclaw/openclaw/issues/34156 | https://api.github.com/repos/openclaw/openclaw/issues/34156 | [Bug]: XML tags leakage in subagent completion events - raw XML visible in user output | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Raw XML tags and template delimiters (e.g., </arg_value>, <|assistant|>) leak into user-visible output during subagent completion events
### Steps to reproduce
1. Configure multiple subagents (builder, reviewer) in openclaw.js... | closed | not_planned | true | 3 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-04T08:11:25Z | 2026-04-22T04:29:59Z | 2026-04-20T04:27:59Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | himiko776-v-6 | 262,384,915 | U_kgDOD6OtEw | User | false |
openclaw/openclaw | 4,020,744,613 | I_kwDOQb6kR87vp7Gl | 34,166 | https://github.com/openclaw/openclaw/issues/34166 | https://api.github.com/repos/openclaw/openclaw/issues/34166 | [Bug]: Gateway failed to start: Error: non-loopback Control UI requires gateway.controlUi.allowedOrigins (set explicit origins), or set gateway.controlUi.dangerouslyAllowHostHeaderOriginFallback=true to use Host-header origin fallback mode | ### Bug type
Regression (worked before, now fails)
### Summary
docker build -t openclaw:local -f Dockerfile .
docker compose run --rm openclaw-cli onboard
docker compose up -d openclaw-gateway
openclaw does not start
### Steps to reproduce
- build images
docker build -t openclaw:local -f Dockerfile .
- settings
-... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T08:25:51Z | 2026-04-22T04:29:58Z | 2026-04-20T04:28:00Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | bin-bin-6 | 23,305,600 | MDQ6VXNlcjIzMzA1NjAw | User | false |
openclaw/openclaw | 4,020,800,885 | I_kwDOQb6kR87vqI11 | 34,180 | https://github.com/openclaw/openclaw/issues/34180 | https://api.github.com/repos/openclaw/openclaw/issues/34180 | [Bug]: unable to carry on | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
node : v22.12.0
pnpm: 10.30.3
Please help me.
I operation `pnpm install` report
```
cript, failed in 1.7sall script...
.../node_modules/node-llama-cpp postinstall$ node ./dist/cli/cli.js postinstall
│ [node-llama-cpp] A prebuilt binary... | closed | not_planned | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-04T08:41:48Z | 2026-04-22T04:29:58Z | 2026-04-20T04:28:02Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | zxzxac110 | 52,234,948 | MDQ6VXNlcjUyMjM0OTQ4 | User | false |
openclaw/openclaw | 4,020,825,312 | I_kwDOQb6kR87vqOzg | 34,188 | https://github.com/openclaw/openclaw/issues/34188 | https://api.github.com/repos/openclaw/openclaw/issues/34188 | [Bug]: Feishu media send returns 230055 when sending mp4 via --media | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
### Command Used
openclaw message send \
--channel feishu \
--account openclaw1 \
--target user:<open_id> \
--media /path/to/video.mp4
### Result
OpenClaw uploads successfully and gets a file_key,
but sending the message fails with ... | closed | not_planned | true | 3 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-04T08:48:22Z | 2026-04-22T04:29:57Z | 2026-04-20T04:28:04Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | jmjl161100 | 207,161,202 | U_kgDODFkHcg | User | false |
openclaw/openclaw | 4,293,081,794 | I_kwDOQb6kR87_4zrC | 69,192 | https://github.com/openclaw/openclaw/issues/69192 | https://api.github.com/repos/openclaw/openclaw/issues/69192 | Hook gateway returns 4xx with no journal entry — breaks diagnosability | ## What's broken
Hook POSTs that fail validation (HTTP 400) leave no trace in the gateway journal. The response body is returned to the caller but nothing is logged on the server side. Makes post-hoc diagnosis of intermittent or production failures impossible — you only have what the (often non-human) caller chose to ... | open | null | false | 1 | [] | [] | 2026-04-20T04:28:05Z | 2026-05-02T14:22:19Z | null | NONE | null | 20260502T150030Z | 2026-05-02T15:00:30Z | bitr8 | 55,131,414 | MDQ6VXNlcjU1MTMxNDE0 | User | false |
openclaw/openclaw | 4,020,827,331 | I_kwDOQb6kR87vqPTD | 34,189 | https://github.com/openclaw/openclaw/issues/34189 | https://api.github.com/repos/openclaw/openclaw/issues/34189 | [Bug]: Control UI / Dashboard returns "Not Found" on Windows (2026.3.x) | ### Bug type
Regression (worked before, now fails)
### Summary
## Control UI / Dashboard returns "Not Found" on Windows (2026.3.x)
### Environment
- **OS:** Windows 10 (10.0.19045, x64)
- **OpenClaw:** 2026.3.2 (installed via npm/pnpm global)
- **Node.js:** v22.19.0
- **Gateway:** Started via `openclaw gateway --p... | closed | not_planned | true | 5 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T08:48:55Z | 2026-04-22T04:29:56Z | 2026-04-20T04:28:06Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | zelixag | 9,192,601 | MDQ6VXNlcjkxOTI2MDE= | User | false |
openclaw/openclaw | 4,020,854,595 | I_kwDOQb6kR87vqV9D | 34,201 | https://github.com/openclaw/openclaw/issues/34201 | https://api.github.com/repos/openclaw/openclaw/issues/34201 | [Bug]: LLM request timed out when switching tools profile from messaging to full | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
I deployed Qwen3.5 locally using vLLM. The model works correctly and normal conversations succeed without any issues.
However, when I modify the openclaw.json configuration file and change the tools profile from messaging to full, the mode... | closed | not_planned | true | 3 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-04T08:55:30Z | 2026-04-22T04:29:56Z | 2026-04-20T04:28:08Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | Jaycee | 62,363,159 | MDQ6VXNlcjYyMzYzMTU5 | User | false |
openclaw/openclaw | 4,020,956,539 | I_kwDOQb6kR87vqu17 | 34,228 | https://github.com/openclaw/openclaw/issues/34228 | https://api.github.com/repos/openclaw/openclaw/issues/34228 | Discord thread sessions reset unexpectedly on each new message (thread starter re-injection) | ## Summary
Discord thread sessions are being silently reset on every new inbound message. The thread starter message gets re-injected as a new user message, which triggers the session reset policy and rotates the transcript to a `.reset.*` file. This causes thread conversations to lose context repeatedly.
## Steps to... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T09:19:27Z | 2026-04-22T04:29:55Z | 2026-04-20T04:28:10Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | sunshineo | 1,072,766 | MDQ6VXNlcjEwNzI3NjY= | User | false |
openclaw/openclaw | 4,020,964,756 | I_kwDOQb6kR87vqw2U | 34,230 | https://github.com/openclaw/openclaw/issues/34230 | https://api.github.com/repos/openclaw/openclaw/issues/34230 | [Bug]: macOS Desktop app starts but window is not visible (v2026.3.2) | ### Bug type
Regression (worked before, now fails)
### Summary
Environment
OpenClaw Desktop: 2026.3.2
macOS: 26.3 (arm64)
Setup: Desktop app on Mac, remote gateway on Raspberry Pi via SSH tunnel
gateway.remote.url on Mac: ws://127.0.0.1:18789 (local tunnel endpoint)
Problem
After updating to 2026.3.2, the OpenClaw ... | closed | not_planned | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T09:21:27Z | 2026-04-22T04:29:54Z | 2026-04-20T04:28:12Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | kkormesser | 27,818,245 | MDQ6VXNlcjI3ODE4MjQ1 | User | false |
openclaw/openclaw | 4,020,982,679 | I_kwDOQb6kR87vq1OX | 34,233 | https://github.com/openclaw/openclaw/issues/34233 | https://api.github.com/repos/openclaw/openclaw/issues/34233 | [Bug]: Local media path is not under an allowed directory: | ### Bug type
Regression (worked before, now fails)
### Summary
我想把我桌面的文件发送到Telegram，但是提示Local media path is not under an allowed directory，我应该如何配置AI可以访问的路径，我想让他整个电脑都可以访问，如果不能访问那就不算一个助手了。我不想去设置什么工作目录。我想设置一个白名单或者一个大的权限。
### Steps to reproduce
All read and write permissions are enabled, but I'm getting the error mess... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T09:25:31Z | 2026-04-22T04:29:53Z | 2026-04-20T04:28:13Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | hudingwen | 36,782,822 | MDQ6VXNlcjM2NzgyODIy | User | false |
openclaw/openclaw | 4,033,295,514 | I_kwDOQb6kR87wZzSa | 37,589 | https://github.com/openclaw/openclaw/issues/37589 | https://api.github.com/repos/openclaw/openclaw/issues/37589 | Avoid recreating unchanged message objects when stripping inline directive tags | ## Problem
`stripInlineDirectiveTagsFromMessageForDisplay` currently recreates the message object and text parts even when no inline directives are present.
## Why this matters
This can cause unnecessary downstream rerenders / diff churn in consumers that rely on object identity to detect actual content changes.
## P... | open | null | false | 3 | [] | [] | 2026-03-06T09:55:05Z | 2026-05-04T09:02:56Z | null | CONTRIBUTOR | null | 20260504T180039Z | 2026-05-04T18:00:39Z | shuofengzhang | 24,763,026 | MDQ6VXNlcjI0NzYzMDI2 | User | false |
openclaw/openclaw | 4,033,313,008 | I_kwDOQb6kR87wZ3jw | 37,591 | https://github.com/openclaw/openclaw/issues/37591 | https://api.github.com/repos/openclaw/openclaw/issues/37591 | [macOS Node] Missing system.run.prepare command prevents system.run execution | # [macOS Node] Missing system.run.prepare command prevents system.run execution
## Problem
The macOS/iOS node software (version 2026.3.2) does not implement the `system.run.prepare` command, which causes `system.run` to fail even when explicitly added to `allowCommands` in the node configuration.
## Environment
- *... | closed | completed | true | 9 | [] | [] | 2026-03-06T09:59:04Z | 2026-04-29T04:33:14Z | 2026-04-27T00:43:19Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | roamerxv | 1,702,265 | MDQ6VXNlcjE3MDIyNjU= | User | false |
openclaw/openclaw | 4,033,324,718 | I_kwDOQb6kR87wZ6au | 37,595 | https://github.com/openclaw/openclaw/issues/37595 | https://api.github.com/repos/openclaw/openclaw/issues/37595 | [Feature Request] Control UI: Show available skills list | ## Description
In the OpenClaw Control UI, add a section to display the agent's available skills.
Currently, users can only view skills via CLI command:
```
openclaw skills list
```
But there's no way to see this in the web UI.
## Expected Behavior
- Show a list of ready/missing skills in the Control UI
- Display ... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T10:01:57Z | 2026-04-27T05:02:32Z | 2026-04-24T19:27:24Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | pitamxa | 152,463,946 | U_kgDOCRZqSg | User | false |
openclaw/openclaw | 4,033,352,217 | I_kwDOQb6kR87waBIZ | 37,605 | https://github.com/openclaw/openclaw/issues/37605 | https://api.github.com/repos/openclaw/openclaw/issues/37605 | [Bug]: Cron jobs lost during gateway restart (race condition in persistence) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Cron jobs created shortly before a gateway restart can be silently lost. The jobs exist in memory and respond to `cron list`, but are not fully persisted to `jobs.json` on disk before the process exits.
### Steps to reproduce
1. Create mu... | closed | completed | true | 4 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-06T10:07:32Z | 2026-04-27T04:55:02Z | 2026-04-24T22:05:22Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Penguin-Life | 265,119,453 | U_kgDOD81m3Q | User | false |
openclaw/openclaw | 4,033,398,832 | I_kwDOQb6kR87waMgw | 37,623 | https://github.com/openclaw/openclaw/issues/37623 | https://api.github.com/repos/openclaw/openclaw/issues/37623 | openai-codex/gpt-5.4 is configurable but not actually supported in runtime (missing/Unknown model) | ### Summary
On OpenClaw `2026.3.2`, `openai-codex/gpt-5.4` can be added to config and appears in `models list`, but it is still treated as `missing` and fails at runtime with `Unknown model` / `HTTP 404`.
This makes it look like GPT-5.4 via `openai-codex` is supported when it is not yet wired through the runtime.
##... | closed | completed | true | 17 | [
"stale"
] | [] | 2026-03-06T10:17:23Z | 2026-04-27T04:55:02Z | 2026-04-24T22:05:25Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mane81 | 75,401,463 | MDQ6VXNlcjc1NDAxNDYz | User | false |
openclaw/openclaw | 4,033,403,490 | I_kwDOQb6kR87waNpi | 37,625 | https://github.com/openclaw/openclaw/issues/37625 | https://api.github.com/repos/openclaw/openclaw/issues/37625 | feat(workspace): lint and auto-fix code fences in markdown at write time | ## Problem
Code stored in markdown workspace files (`MEMORY.md`, `AGENTS.md`, skill files, etc.) frequently appears without proper fencing, or with fencing that's missing a language tag. This causes:
- **Rendering failures** in Discord/Telegram where unfenced code is displayed as plain text
- **Syntax highlighting ga... | open | null | false | 2 | [] | [] | 2026-03-06T10:18:32Z | 2026-05-04T10:02:34Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | aronchick | 51,317 | MDQ6VXNlcjUxMzE3 | User | false |
openclaw/openclaw | 4,033,403,873 | I_kwDOQb6kR87waNvh | 37,626 | https://github.com/openclaw/openclaw/issues/37626 | https://api.github.com/repos/openclaw/openclaw/issues/37626 | [feishu plugin] wiki spaceNode.list 分页不生效，只返回前20条记录 | ## 问题描述
飞书插件的 `feishu_wiki` 工具在调用 `nodes` action 时，无法获取完整的子节点列表。飞书 API 默认只返回前 20 条记录，即使添加了分页支持代码也无法正确获取所有页面。
## 环境信息
- 插件版本: \`@openclaw/feishu\` (local)
- 插件路径: \`~/.openclaw/extensions/feishu\`
- Node.js: v25.6.0
- OS: macOS (Darwin 25.3.0 arm64)
## 复现步骤
1. 使用 \`feishu_wiki\` 工具调用 \`nodes\` action 获取一个有超过 20 个子节... | open | null | false | 5 | [] | [] | 2026-03-06T10:18:38Z | 2026-05-04T09:50:00Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | ritou11 | 17,564,328 | MDQ6VXNlcjE3NTY0MzI4 | User | false |
openclaw/openclaw | 4,033,419,557 | I_kwDOQb6kR87waRkl | 37,630 | https://github.com/openclaw/openclaw/issues/37630 | https://api.github.com/repos/openclaw/openclaw/issues/37630 | feat(cron): hierarchical job store — one file per cron job in cron/jobs/ directory | ## Problem
Cron jobs are stored in a single monolithic `~/.openclaw/cron/jobs.json` file. As users accumulate jobs, this becomes:
- **Hard to edit** — finding one job in a large JSON array requires scrolling/searching
- **Merge-unfriendly** — any concurrent write touches the same file, creating conflicts or races
- *... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-06T10:22:29Z | 2026-04-28T04:43:46Z | 2026-04-26T01:31:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | aronchick | 51,317 | MDQ6VXNlcjUxMzE3 | User | false |
openclaw/openclaw | 4,033,432,805 | I_kwDOQb6kR87waUzl | 37,633 | https://github.com/openclaw/openclaw/issues/37633 | https://api.github.com/repos/openclaw/openclaw/issues/37633 | [Bug]: lark/feishu error plugin status with unresolved appsecret(using env) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
It stays error status.
Error logs
```
[plugins] feishu failed during register from /home/xxx/.nvm/versions/node/v22.22.0/lib/node_modules/openclaw/extensions/feishu/index.ts: Error: channels.feishu.appSecret: unresolved SecretRef "env:defau... | closed | completed | true | 6 | [
"bug",
"bug:behavior"
] | [] | 2026-03-06T10:25:44Z | 2026-04-30T04:51:00Z | 2026-04-25T05:38:30Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | CodeMasterYi | 10,430,162 | MDQ6VXNlcjEwNDMwMTYy | User | false |
openclaw/openclaw | 4,033,434,403 | I_kwDOQb6kR87waVMj | 37,634 | https://github.com/openclaw/openclaw/issues/37634 | https://api.github.com/repos/openclaw/openclaw/issues/37634 | sandbox: keep workspaceAccess none workspaces writable | ## Summary
When `agents.defaults.sandbox.workspaceAccess` is set to `"none"`, OpenClaw gives each sandboxed session its own isolated workspace under `~/.openclaw/sandboxes/...`, which matches the expected isolation model. However, that isolated `/workspace` ends up mounted read-only, so tools that need to write inside ... | open | null | false | 7 | [] | [] | 2026-03-06T10:26:05Z | 2026-05-04T09:03:01Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | whyuds | 6,570,978 | MDQ6VXNlcjY1NzA5Nzg= | User | false |
openclaw/openclaw | 4,033,441,849 | I_kwDOQb6kR87waXA5 | 37,635 | https://github.com/openclaw/openclaw/issues/37635 | https://api.github.com/repos/openclaw/openclaw/issues/37635 | models list --all omits config-injected models from models.providers | ### Summary
`openclaw models list --all` does not include provider models that are injected through `models.providers.<provider>.models` when those models are not already present in the underlying `pi-ai` registry.
This makes the `--all` view incomplete for forward-compat / shadow-registry setups, even when OpenClaw ... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T10:27:38Z | 2026-04-27T05:02:31Z | 2026-04-24T19:27:33Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Sapientropic | 194,952,804 | U_kgDOC56-ZA | User | false |
openclaw/openclaw | 4,033,478,313 | I_kwDOQb6kR87waf6p | 37,645 | https://github.com/openclaw/openclaw/issues/37645 | https://api.github.com/repos/openclaw/openclaw/issues/37645 | patchToolSchemaForClaudeCompatibility empties required[] — non-Claude models send empty arguments | ## Bug
`patchToolSchemaForClaudeCompatibility()` in `pi-coding-agent` (bundled in `pi-embedded-*.js`) adds Claude-compatible aliases for tool parameters:
- `path` → `file_path`
- `oldText` → `old_string`
- `newText` → `new_string`
However, it **removes** the original param from the `required` array without adding th... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-06T10:35:32Z | 2026-04-27T04:55:01Z | 2026-04-24T22:05:29Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | koburafab | 26,431,644 | MDQ6VXNlcjI2NDMxNjQ0 | User | false |
openclaw/openclaw | 4,033,483,102 | I_kwDOQb6kR87wahFe | 37,647 | https://github.com/openclaw/openclaw/issues/37647 | https://api.github.com/repos/openclaw/openclaw/issues/37647 | [Bug]: TUI doesn't display assistant replies, but webchat works fine | ### Bug type
Regression (worked before, now fails)
### Summary
Assistant messages are not displayed in the terminal TUI, but they work correctly in the webchat interface (http://127.0.0.1:18789/chat).
- OpenClaw version: 2026.3.2
- Platform: Linux (systemd service)
- TUI command: openclaw tui
- Gateway URL: ws://127... | closed | completed | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-06T10:36:41Z | 2026-04-28T04:43:45Z | 2026-04-26T01:32:03Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dechace5-hash | 265,997,837 | U_kgDOD9rODQ | User | false |
openclaw/openclaw | 4,033,485,167 | I_kwDOQb6kR87wahlv | 37,651 | https://github.com/openclaw/openclaw/issues/37651 | https://api.github.com/repos/openclaw/openclaw/issues/37651 | Feature request: Add lightweight shell payload type for cron jobs | # GitHub Issue Draft — openclaw/openclaw
---
**Title:** Feature request: Add lightweight `shell` payload type for cron jobs
---
## Problem
Currently, cron jobs support two payload types:
- `systemEvent` — injects text into the main session (requires LLM)
- `agentTurn` — runs in an isolated session (requires LLM, ... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-06T10:37:14Z | 2026-04-28T04:43:44Z | 2026-04-26T01:32:10Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | cnJasonZ | 3,736,986 | MDQ6VXNlcjM3MzY5ODY= | User | false |
openclaw/openclaw | 4,033,498,869 | I_kwDOQb6kR87wak71 | 37,655 | https://github.com/openclaw/openclaw/issues/37655 | https://api.github.com/repos/openclaw/openclaw/issues/37655 | Cron Jobs page: New Job panel overlaps Jobs list columns | ## Bug
On the **Cron Jobs** page in the Control UI, the "New Job" panel on the right overlaps and covers the Jobs list table columns (Enabled, Schedule, Last run, Sort, Direction, Reset).
## Screenshot
The Jobs list table and the New Job form are rendered side-by-side, but the New Job panel clips over the right-side... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-06T10:40:41Z | 2026-04-27T05:02:31Z | 2026-04-24T19:27:40Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | floatinggball-design | 262,259,579 | U_kgDOD6HDew | User | false |
openclaw/openclaw | 4,033,525,515 | I_kwDOQb6kR87warcL | 37,659 | https://github.com/openclaw/openclaw/issues/37659 | https://api.github.com/repos/openclaw/openclaw/issues/37659 | docs: improve safe-mode agent instructions for token updates and OAuth re-auth | ## Summary
During a safe-mode recovery session, two documentation gaps caused extended debugging:
### Issue 1: Token Updates - Missing habitat-parsed.env
**Current state:** TOOLS.md mentions `/etc/habitat-parsed.env` in Key Paths but doesn't explain its role.
**Problem:** When Discord bot tokens are regenerated, th... | closed | not_planned | true | 3 | [] | [] | 2026-03-06T10:47:16Z | 2026-04-29T04:35:51Z | 2026-04-26T14:55:39Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | dfrysinger | 1,424,648 | MDQ6VXNlcjE0MjQ2NDg= | User | false |
openclaw/openclaw | 4,033,541,051 | I_kwDOQb6kR87wavO7 | 37,661 | https://github.com/openclaw/openclaw/issues/37661 | https://api.github.com/repos/openclaw/openclaw/issues/37661 | [Bug]: LLM streaming output infinite loop - same phrase repeated 40+ times | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When using subagent spawns with embedded runs, the LLM output can enter an infinite loop state where the same phrase is repeated dozens of times in a single message, and multiple duplicate messages are sent to the user channel.
### Steps t... | open | null | false | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-06T10:51:08Z | 2026-05-04T06:15:00Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | himiko776-v-6 | 262,384,915 | U_kgDOD6OtEw | User | false |
openclaw/openclaw | 4,033,551,111 | I_kwDOQb6kR87waxsH | 37,667 | https://github.com/openclaw/openclaw/issues/37667 | https://api.github.com/repos/openclaw/openclaw/issues/37667 | Feature: Cross-channel session continuity for same-user DMs | ## Problem
When a single user talks to the same agent across multiple channels (e.g., WhatsApp → webchat → Telegram), each channel creates an isolated session. The agent on channel B has zero awareness of the conversation that just happened on channel A.
From the user's perspective, they're talking to **one assistant... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T10:53:28Z | 2026-04-27T05:02:30Z | 2026-04-24T19:27:51Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | clawdiri-ai | 266,047,309 | U_kgDOD9uPTQ | User | false |
openclaw/openclaw | 4,033,605,638 | I_kwDOQb6kR87wa_AG | 37,675 | https://github.com/openclaw/openclaw/issues/37675 | https://api.github.com/repos/openclaw/openclaw/issues/37675 | [Bug] Webchat: input field content lost when cron/subagent announce messages arrive | ## Bug type
Behavior bug (incorrect output/state without crash)
## Summary
When using the webchat Control UI and typing a message, if a cron job announce or subagent completion announce arrives, the chat input field is cleared/reset — losing whatever the user was typing. This is disruptive during active conversations ... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-06T11:06:46Z | 2026-04-27T05:02:29Z | 2026-04-24T19:28:00Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | willitude | 244,889,796 | U_kgDODpi4xA | User | false |
openclaw/openclaw | 4,033,641,320 | I_kwDOQb6kR87wbHto | 37,696 | https://github.com/openclaw/openclaw/issues/37696 | https://api.github.com/repos/openclaw/openclaw/issues/37696 | unclosed <think> tags silently drop entire response on all channels | ## Bug
When a model emits `<think>` without a closing `</think>` (common with Qwen 3.5, observed intermittently on short casual responses), `stripReasoningTagsFromText()` in **strict mode** discards ALL text after the unclosed tag. The user receives nothing — no error, no partial response, just silence.
## Impact
- ... | closed | completed | true | 4 | [] | [] | 2026-03-06T11:14:11Z | 2026-04-30T04:34:27Z | 2026-04-27T11:45:23Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | druide67 | 212,749,853 | U_kgDODK5OHQ | User | false |
openclaw/openclaw | 4,033,678,467 | I_kwDOQb6kR87wbQyD | 37,701 | https://github.com/openclaw/openclaw/issues/37701 | https://api.github.com/repos/openclaw/openclaw/issues/37701 | macOS app crashes on Talk mode launch (TalkOverlayView exclusivity violation) | ## Crash
The macOS companion app (v2026.3.2, build 2026030290) crashes immediately when Talk mode is enabled. The app never shows the Talk overlay — it aborts on every attempt.
## Environment
- macOS 26.2 (25C56), Apple Silicon (Mac16,10 — Mac mini)
- OpenClaw macOS app 2026.3.2 (2026030290)
- Gateway 2026.3.2 runnin... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-06T11:23:16Z | 2026-04-27T05:02:27Z | 2026-04-24T19:28:12Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | floatinggball-design | 262,259,579 | U_kgDOD6HDew | User | false |
openclaw/openclaw | 4,033,682,305 | I_kwDOQb6kR87wbRuB | 37,702 | https://github.com/openclaw/openclaw/issues/37702 | https://api.github.com/repos/openclaw/openclaw/issues/37702 | Outbound Telegram dedup resets after context compaction — causes duplicate messages | ## Problem
After context compaction, the assistant's outbound message dedup guard resets because it only tracks duplicates within a single `assistantMessageIndex`. When the model re-generates text from the compaction summary as new output, it gets delivered to Telegram as a duplicate message.
## Reproduction
1. Have... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T11:24:16Z | 2026-04-27T05:02:27Z | 2026-04-24T19:28:16Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Mquarmoc | 179,647,526 | U_kgDOCrU0Jg | User | false |
openclaw/openclaw | 4,033,692,040 | I_kwDOQb6kR87wbUGI | 37,705 | https://github.com/openclaw/openclaw/issues/37705 | https://api.github.com/repos/openclaw/openclaw/issues/37705 | [Bug]: Gateway Launch Failure: Label cannot be longer than 63 bytes | ### Bug type
Crash (process/app exits or hangs)
### Summary
`[openclaw] Uncaught exception: AssertionError [ERR_ASSERTION]: Label cannot be longer than 63 bytes (app-41627eae5842473f9e05f139ea307277-7f9477f4d6-lqqzf (OpenClaw))
at DNSLabelCoder.getLabelLength (envs/openclaw/lib/node_modules/openclaw/node_modules... | closed | completed | true | 4 | [
"bug",
"bug:crash"
] | [
"vincentkoc"
] | 2026-03-06T11:26:45Z | 2026-04-30T04:33:18Z | 2026-04-27T20:19:28Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | LHucass | 123,493,995 | U_kgDOB1xeaw | User | false |
openclaw/openclaw | 4,033,693,984 | I_kwDOQb6kR87wbUkg | 37,706 | https://github.com/openclaw/openclaw/issues/37706 | https://api.github.com/repos/openclaw/openclaw/issues/37706 | [Bug]: Feishu/Lark group @all should not count as bot mention | ## Summary
In Feishu/Lark group chats, messages that only contain `@all` / `@_all` are currently treated as if the bot was explicitly mentioned.
This causes the bot to reply to broadcast notifications even when no one intended to talk to the bot.
## Affected version
- OpenClaw `2026.3.2`
- Official `@openclaw/feish... | closed | completed | true | 2 | [
"clawsweeper"
] | [] | 2026-03-06T11:27:17Z | 2026-04-30T04:33:40Z | 2026-04-27T09:10:19Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | abca12a | 35,697,843 | MDQ6VXNlcjM1Njk3ODQz | User | false |
openclaw/openclaw | 4,033,705,068 | I_kwDOQb6kR87wbXRs | 37,711 | https://github.com/openclaw/openclaw/issues/37711 | https://api.github.com/repos/openclaw/openclaw/issues/37711 | [Bug]: Hostinger Docker template deploys successfully, but OpenClaw web UI stays black due to /assets/index-*.js ERR_CONTENT_LENGTH_MISMATCH | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Fresh Hostinger template deployments start successfully, but the OpenClaw web UI stays black because the main /assets/index-*.js request fails with ERR_CONTENT_LENGTH_MISMATCH.
### Steps to reproduce
1. Deploy OpenClaw on a fresh Hostinge... | open | null | false | 4 | [
"bug",
"bug:behavior"
] | [] | 2026-03-06T11:30:08Z | 2026-05-04T06:15:01Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | DOU0319 | 130,894,211 | U_kgDOB81Jgw | User | false |
openclaw/openclaw | 4,033,714,474 | I_kwDOQb6kR87wbZkq | 37,716 | https://github.com/openclaw/openclaw/issues/37716 | https://api.github.com/repos/openclaw/openclaw/issues/37716 | [Bug] 心跳间隔无法配置，实际频率远高于文档说明的 30 分钟 | # [Bug] 心跳间隔无法配置，实际频率远高于文档说明的 30 分钟
## 问题描述
根据官方文档 (FAQ)，心跳间隔应该是 30 分钟一次，但实际使用中发现心跳频率约为 **5 分钟一次**，且无法通过配置文件修改。
## 环境信息
- **OpenClaw 版本**: 2026.3.1
- **操作系统**: Windows 11 25H2
- **Node 版本**: 24.14.0
- **安装方式**: npm
- **Channel**: stable
## 复现步骤
1. 安装 OpenClaw 并完成初始化
2. 启动 Gateway
3. 观察日志中的心跳触发记录
4. 尝试在 `openclaw.... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T11:32:04Z | 2026-04-27T05:02:26Z | 2026-04-24T19:28:21Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jonah791 | 175,467,022 | U_kgDOCnVqDg | User | false |
openclaw/openclaw | 4,033,717,899 | I_kwDOQb6kR87wbaaL | 37,718 | https://github.com/openclaw/openclaw/issues/37718 | https://api.github.com/repos/openclaw/openclaw/issues/37718 | [Bug]: Gateway health check fails when using secrets reference format for channel credentials (cfg?.appId?.trim is not a function) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Description
When channel credentials (Feishu appId/appSecret, WeCom bot.token/agent.corpId, etc.) are configured using secrets reference format, the gateway health check and configuration parsing fail with TypeError: cfg?.appId?.trim is not... | closed | completed | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-06T11:32:43Z | 2026-04-28T04:43:43Z | 2026-04-26T01:32:29Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hao65103940 | 33,945,037 | MDQ6VXNlcjMzOTQ1MDM3 | User | false |
openclaw/openclaw | 4,033,809,322 | I_kwDOQb6kR87wbwuq | 37,743 | https://github.com/openclaw/openclaw/issues/37743 | https://api.github.com/repos/openclaw/openclaw/issues/37743 | [Bug]: Browser disconnects on tradingview | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Environment:
- OpenClaw version: 2026.2.25
- OS: Ubuntu 24.04 LTS in Oracle VirtualBox
- Chrome extension: OpenClaw Copilot
- Gateway: running as systemd service on port 18789
### Steps to reproduce
ssue 1 — Browser Extension:
Steps to... | closed | completed | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-06T11:52:39Z | 2026-04-27T05:02:25Z | 2026-04-24T19:28:26Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | jthompsonenterprise-ctrl | 266,058,600 | U_kgDOD9u7aA | User | false |
openclaw/openclaw | 4,033,827,323 | I_kwDOQb6kR87wb1H7 | 37,746 | https://github.com/openclaw/openclaw/issues/37746 | https://api.github.com/repos/openclaw/openclaw/issues/37746 | [Feature]: Sandboxed skill functions for remote heartbeat instructions | ## Problem
Skills that integrate with live platforms (forums, chat networks, social agents) need a way to fetch dynamic instructions at runtime — not just static behavior baked into the skill at install time. The canonical example is a `heartbeat.md` pattern where an agent periodically fetches a URL and follows whatev... | open | null | false | 2 | [] | [] | 2026-03-06T11:56:53Z | 2026-05-04T06:15:02Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | AndrewAndrewsen | 34,192,400 | MDQ6VXNlcjM0MTkyNDAw | User | false |
openclaw/openclaw | 4,033,830,574 | I_kwDOQb6kR87wb16u | 37,748 | https://github.com/openclaw/openclaw/issues/37748 | https://api.github.com/repos/openclaw/openclaw/issues/37748 | skill-creator: make .skill package file order deterministic | ## Summary
`skills/skill-creator/scripts/package_skill.py` currently walks files via `Path.rglob("*")` without enforcing order. Traversal order can differ across filesystems/environments, which makes generated `.skill` archives non-deterministic.
## Proposed improvement
Sort files by relative path before writing archi... | open | null | false | 2 | [] | [] | 2026-03-06T11:57:44Z | 2026-05-04T06:03:26Z | null | CONTRIBUTOR | null | 20260504T090105Z | 2026-05-04T09:01:05Z | shuofengzhang | 24,763,026 | MDQ6VXNlcjI0NzYzMDI2 | User | false |
openclaw/openclaw | 4,033,836,340 | I_kwDOQb6kR87wb3U0 | 37,749 | https://github.com/openclaw/openclaw/issues/37749 | https://api.github.com/repos/openclaw/openclaw/issues/37749 | Onboard should auto-approve local node host pairing or prompt user | ## Summary
After running `openclaw onboard`, the local node host service (`ai.openclaw.node` LaunchAgent) is installed and started automatically, but its device pairing request (`role: node`) is never approved. This causes the node host to enter an infinite reconnect loop with `pairing required` errors in the gateway ... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T11:59:14Z | 2026-04-27T05:02:24Z | 2026-04-24T19:28:30Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | AI-Kid1412 | 263,868,266 | U_kgDOD7pPag | User | false |
openclaw/openclaw | 4,033,859,591 | I_kwDOQb6kR87wb9AH | 37,752 | https://github.com/openclaw/openclaw/issues/37752 | https://api.github.com/repos/openclaw/openclaw/issues/37752 | [Bug]: Discord channel reconnect timeout triggers full Gateway restart | ### Bug type
Regression (worked before, now fails)
### Summary
Discord WebSocket 断开后重连超时（5 分钟），triggerForceStop() 错误传播到 Gateway 主进程，导致整个 Gateway 重启，影响所有其他通道（如飞书）
### Steps to reproduce
1. 启用 Discord 通道（配置 token）
2. 等待 Discord WebSocket 断开（code 1006，网络波动导致）
3. Discord 尝试重连
4. 等待 5 分钟（RECONNECT_STALL_TIMEOUT_MS）
5. ... | closed | completed | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-06T12:05:19Z | 2026-04-27T05:02:22Z | 2026-04-24T19:28:35Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | YHlorra | 111,432,771 | U_kgDOBqRUQw | User | false |
openclaw/openclaw | 4,033,891,723 | I_kwDOQb6kR87wcE2L | 37,757 | https://github.com/openclaw/openclaw/issues/37757 | https://api.github.com/repos/openclaw/openclaw/issues/37757 | System prompt typo: IDENTITITY.md vs IDENTITY.md | **Description**: The system prompt display shows `IDENTITITY.md` (with double 'I') in the "Workspace Files (injected)" section, but the actual file is correctly named `IDENTITY.md` and loads fine.
**Impact**: Cosmetic bug only - the file loads correctly, but the typo in the generated system prompt could cause confusio... | closed | not_planned | true | 27 | [
"stale"
] | [] | 2026-03-06T12:12:34Z | 2026-04-27T05:02:19Z | 2026-04-24T19:28:39Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | stormhierta | 51,702,891 | MDQ6VXNlcjUxNzAyODkx | User | false |
openclaw/openclaw | 4,033,895,307 | I_kwDOQb6kR87wcFuL | 37,758 | https://github.com/openclaw/openclaw/issues/37758 | https://api.github.com/repos/openclaw/openclaw/issues/37758 | [Feishu] Multi-agent routing still broken - requesting ETA and workaround | ## Summary
We need to run **multiple independent agents on separate Feishu bots** (each with its own personality, memory, and workspace). However, this is currently broken due to the session path resolution bug described in #15246 and the routing issues in #8692.
## Our Setup
- OpenClaw: v2026.2.26
- Channel: Feishu... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-06T12:13:17Z | 2026-04-27T05:02:17Z | 2026-04-24T19:28:44Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | smellyday | 558,941 | MDQ6VXNlcjU1ODk0MQ== | User | false |
openclaw/openclaw | 4,033,919,415 | I_kwDOQb6kR87wcLm3 | 37,763 | https://github.com/openclaw/openclaw/issues/37763 | https://api.github.com/repos/openclaw/openclaw/issues/37763 | Slack: agent identity (username/icon_emoji) not applied to messages | ## Summary
Agent identity configured in `agents.list[].identity` is not being applied to Slack messages. Messages still show the bot's default display name instead of the configured identity name.
## Config
```json
{
"agents": {
"list": [
{
"id": "main",
"identity": {
"name": "Sti... | closed | not_planned | true | 2 | [] | [] | 2026-03-06T12:19:06Z | 2026-04-29T04:38:05Z | 2026-04-26T10:02:32Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | jassoulane | 265,021,824 | U_kgDOD8vpgA | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.