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 | 4,029,137,828 | I_kwDOQb6kR87wJ8Ok | 36,385 | https://github.com/openclaw/openclaw/issues/36385 | https://api.github.com/repos/openclaw/openclaw/issues/36385 | skill-creator: quick validator should reject empty name/description | ## Problem
`skills/skill-creator/scripts/quick_validate.py` treated `name` and `description` as required keys, but still accepted empty/whitespace-only values. This lets invalid skill metadata pass validation.
## Proposed improvement
Require non-empty values for both `name` and `description` in SKILL.md frontmatter an... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T15:54:45Z | 2026-04-25T04:18:07Z | 2026-04-22T04:33:55Z | CONTRIBUTOR | null | 20260425T060021Z | 2026-04-25T06:00:21Z | shuofengzhang | 24,763,026 | MDQ6VXNlcjI0NzYzMDI2 | User | false |
openclaw/openclaw | 4,029,093,584 | I_kwDOQb6kR87wJxbQ | 36,377 | https://github.com/openclaw/openclaw/issues/36377 | https://api.github.com/repos/openclaw/openclaw/issues/36377 | [Bug]: Mem0 OSS Plugin unusable on OpenClaw 2026.3.x due to jiti/sqlite3 binding regression | ### Bug type
Regression (worked before, now fails)
### Summary
The following problem has already been reported in [#31676](https://github.com/openclaw/openclaw/issues/31676) and [#31677](https://github.com/openclaw/openclaw/issues/31677). But since both have been closed (one probably by mistake, because 31676 has be... | closed | not_planned | true | 5 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T15:47:29Z | 2026-04-25T04:18:07Z | 2026-04-22T04:33:54Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | TRON4R | 70,440,224 | MDQ6VXNlcjcwNDQwMjI0 | User | false |
openclaw/openclaw | 4,028,865,168 | I_kwDOQb6kR87wI5qQ | 36,326 | https://github.com/openclaw/openclaw/issues/36326 | https://api.github.com/repos/openclaw/openclaw/issues/36326 | Hook transform `model` override ignored at runtime | ## Description
When a hook transform returns a `model` field (e.g., `model: "anthropic-proxy/claude-sonnet-4-6"`), the model override is correctly merged into the action via `mergeAction` in `hooks-mapping.ts`, passed through `dispatchAgentHook`, and stored in `job.payload.model`. However, at runtime the isolated agen... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T15:11:21Z | 2026-04-25T04:18:08Z | 2026-04-22T04:33:52Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | cioclawcode | 261,991,534 | U_kgDOD52sbg | User | false |
openclaw/openclaw | 4,028,862,179 | I_kwDOQb6kR87wI47j | 36,325 | https://github.com/openclaw/openclaw/issues/36325 | https://api.github.com/repos/openclaw/openclaw/issues/36325 | Hook `deliver: false` still injects system event into main session | ## Description
When a hook mapping sets `deliver: false`, the delivery announcement is correctly suppressed, but a system event is still injected into the main session via `enqueueSystemEvent`. This causes the main session to receive "Hook <name>: <summary>" messages for every hook completion, even when the user expli... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T15:10:52Z | 2026-04-25T04:18:09Z | 2026-04-22T04:33:50Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | cioclawcode | 261,991,534 | U_kgDOD52sbg | User | false |
openclaw/openclaw | 4,028,774,788 | I_kwDOQb6kR87wIjmE | 36,301 | https://github.com/openclaw/openclaw/issues/36301 | https://api.github.com/repos/openclaw/openclaw/issues/36301 | [Bug]: Gateway UI/Service Version Discrepancy After NPM Global Update | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
I've observed that after updating OpenClaw globally using `npm install -g openclaw@latest`, the OpenClaw Gateway service (as reported in the control UI's "Instances" page) continues to display an older version number, even though `openclaw ... | closed | not_planned | true | 4 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-05T14:57:26Z | 2026-04-25T04:18:10Z | 2026-04-22T04:33:47Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | pshotts | 40,694,049 | MDQ6VXNlcjQwNjk0MDQ5 | User | false |
openclaw/openclaw | 4,028,665,999 | I_kwDOQb6kR87wIJCP | 36,270 | https://github.com/openclaw/openclaw/issues/36270 | https://api.github.com/repos/openclaw/openclaw/issues/36270 | [Bug]: run error: HTTP 401: Invalid Authentication | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Got a error right after install the app.
<img width="2462" height="1162" alt="Image" src="https://github.com/user-attachments/assets/3e219edf-dff5-4a74-a81b-5e7471b57e36" />
Here's `gateway.err.log`
```
2026-03-05T14:32:06.771Z [gateway]... | closed | not_planned | true | 3 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-05T14:40:35Z | 2026-04-25T04:18:10Z | 2026-04-22T04:33:45Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | wuwen1030 | 3,377,567 | MDQ6VXNlcjMzNzc1Njc= | User | false |
openclaw/openclaw | 4,028,640,719 | I_kwDOQb6kR87wIC3P | 36,267 | https://github.com/openclaw/openclaw/issues/36267 | https://api.github.com/repos/openclaw/openclaw/issues/36267 | status inconsistency: Bncr shown as SETUP/not linked while deep/json are linked | ## Bug: `openclaw status` top Channels row shows Bncr `SETUP/not linked` while `status --deep` and `health --json` show linked
### Environment
- OpenClaw version: `2026.3.2`
- Host: Debian 13
- Channel: custom plugin `bncr`
- Bncr accountId in config: non-default (e.g. `Primary`)
### What happens
`openclaw status` sh... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T14:36:49Z | 2026-04-25T04:18:11Z | 2026-04-22T04:33:43Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | xmoxmo | 112,841,961 | U_kgDOBrnU6Q | User | false |
openclaw/openclaw | 4,028,624,320 | I_kwDOQb6kR87wH-3A | 36,263 | https://github.com/openclaw/openclaw/issues/36263 | https://api.github.com/repos/openclaw/openclaw/issues/36263 | fix(cron): payload.fallbacks not persisted via cron.update API | ## Bug
`payload.fallbacks` (array of fallback model strings for `agentTurn` cron jobs) is **read at runtime** (`runWithModelFallback` at line ~4152 of `gateway-cli-vk3t7zJU.js`) but **not persisted** when updating a cron job via the `cron.update` API.
This is the exact same class of bug as #31425 (`lightContext` not ... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T14:34:18Z | 2026-04-25T04:18:12Z | 2026-04-22T04:33:42Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | CountofGlamorgan | 59,939,121 | MDQ6VXNlcjU5OTM5MTIx | User | false |
openclaw/openclaw | 4,028,575,629 | I_kwDOQb6kR87wHy-N | 36,256 | https://github.com/openclaw/openclaw/issues/36256 | https://api.github.com/repos/openclaw/openclaw/issues/36256 | [Bug]: Telegram voice updates are received, but no STT/transcription pipeline executes (OpenClaw 2026.3.2) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Environment:
OpenClaw: 2026.3.2
OS: macOS 26.3 (arm64)
Node: 25.7.0
Channel: Telegram bot DM (polling mode), channel status OK
Config:
tools.media.audio.enabled = true
skills.entries.openai-whisper-api.apiKey configured
Gateway healthy/rea... | closed | not_planned | true | 3 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-05T14:26:58Z | 2026-04-25T04:18:13Z | 2026-04-22T04:33:40Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | jasonkr-hue | 265,825,398 | U_kgDOD9gsdg | User | false |
openclaw/openclaw | 4,028,529,605 | I_kwDOQb6kR87wHnvF | 36,248 | https://github.com/openclaw/openclaw/issues/36248 | https://api.github.com/repos/openclaw/openclaw/issues/36248 | sessions_send in group chat: announce delivery fails with 400 error and causes session to hang | ## Bug Description
When using `sessions_send` to have another agent send a message in a group chat, the announce delivery frequently fails with `AxiosError: Request failed with status code 400`. This causes the session to hang and become unresponsive.
## Steps to Reproduce
1. Have multiple agents (main, life, infoma... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T14:19:50Z | 2026-04-25T04:18:14Z | 2026-04-22T04:33:38Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | hadiliu | 1,926,042 | MDQ6VXNlcjE5MjYwNDI= | User | false |
openclaw/openclaw | 4,028,526,295 | I_kwDOQb6kR87wHm7X | 36,247 | https://github.com/openclaw/openclaw/issues/36247 | https://api.github.com/repos/openclaw/openclaw/issues/36247 | sessions_send in group chat: announce delivery fails with 400 error and causes session to hang | ## Bug Description
When using `sessions_send` to have another agent send a message in a group chat, the announce delivery frequently fails with `AxiosError: Request failed with status code 400`. This causes the session to hang and become unresponsive.
## Steps to Reproduce
1. Have multiple agents (main, life, infoma... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T14:19:21Z | 2026-04-25T04:18:14Z | 2026-04-22T04:33:36Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | hadiliu | 1,926,042 | MDQ6VXNlcjE5MjYwNDI= | User | false |
openclaw/openclaw | 4,028,408,815 | I_kwDOQb6kR87wHKPv | 36,229 | https://github.com/openclaw/openclaw/issues/36229 | https://api.github.com/repos/openclaw/openclaw/issues/36229 | Bug: compaction corrupts thinking block signatures, causing unrecoverable session failure | ## Summary
When a long-running session with `thinking: high` (or any extended thinking mode) approaches context limits and triggers auto-compaction, the message sanitization pipeline modifies thinking block signatures on retained messages. This causes Anthropic to reject all subsequent API calls with:
```
thinking/re... | closed | not_planned | true | 10 | [
"stale"
] | [] | 2026-03-05T14:00:58Z | 2026-04-25T04:18:15Z | 2026-04-22T04:33:35Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | zylos01 | 258,754,527 | U_kgDOD2xH3w | User | false |
openclaw/openclaw | 4,028,366,132 | I_kwDOQb6kR87wG_00 | 36,221 | https://github.com/openclaw/openclaw/issues/36221 | https://api.github.com/repos/openclaw/openclaw/issues/36221 | [Bug] Messages deleted/replaced mid-display when switching to fallback model on context limit | ## Problem
When a conversation reaches the token limit and OpenClaw switches to a fallback AI model, messages that have already been posted and displayed in the UI are being deleted or replaced mid-display.
This means:
- A message appears in the chat
- While reading it, it disappears and is replaced by the next mess... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T13:53:29Z | 2026-04-25T04:18:16Z | 2026-04-22T04:33:33Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | JaguarsMark2 | 234,004,174 | U_kgDODfKezg | User | false |
openclaw/openclaw | 4,028,365,243 | I_kwDOQb6kR87wG_m7 | 36,220 | https://github.com/openclaw/openclaw/issues/36220 | https://api.github.com/repos/openclaw/openclaw/issues/36220 | Feature: subagent skill inheritance | ## Problem
When spawning subagents via `sessions_spawn`, the subagent has no access to the parent session's skill list. It only receives the task prompt, so any skill-specific workflows must be manually transcribed into the spawn prompt — which leads to omissions and inconsistencies.
## Example
Parent session has `bid... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T13:53:19Z | 2026-04-25T04:18:17Z | 2026-04-22T04:33:31Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | zacjiang | 15,880,609 | MDQ6VXNlcjE1ODgwNjA5 | User | false |
openclaw/openclaw | 4,028,229,398 | I_kwDOQb6kR87wGecW | 36,190 | https://github.com/openclaw/openclaw/issues/36190 | https://api.github.com/repos/openclaw/openclaw/issues/36190 | [Feature]: agents.defaults.systemPromptFile — file-based system prompt injection from config | ## Summary
Allow operators to inject a custom system prompt from an external file path specified in config, applied to all sessions, and not editable by the agent.
## Problem to solve
Operators who run openclaw-based agents need to enforce behavioral instructions (guardrails, persona, operational rules) via the syst... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T13:27:02Z | 2026-04-25T04:18:18Z | 2026-04-22T04:33:29Z | CONTRIBUTOR | null | 20260425T060021Z | 2026-04-25T06:00:21Z | carrotRakko | 24,588,751 | MDQ6VXNlcjI0NTg4NzUx | User | false |
openclaw/openclaw | 4,028,221,548 | I_kwDOQb6kR87wGchs | 36,185 | https://github.com/openclaw/openclaw/issues/36185 | https://api.github.com/repos/openclaw/openclaw/issues/36185 | message tool uses default media roots instead of agent-scoped roots | ## Bug
The `message` tool's media path validation uses `getDefaultMediaLocalRoots()` instead of `getAgentScopedMediaLocalRoots()`, so files under the agent's configured workspace directory are rejected with `LocalMediaAccessError: path-not-allowed`.
## Reproduction
- Agent workspace: `/home/rav/clawd` (configured in... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T13:25:29Z | 2026-04-25T04:18:18Z | 2026-04-22T04:33:27Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | ravz | 1,230,289 | MDQ6VXNlcjEyMzAyODk= | User | false |
openclaw/openclaw | 4,028,217,141 | I_kwDOQb6kR87wGbc1 | 36,182 | https://github.com/openclaw/openclaw/issues/36182 | https://api.github.com/repos/openclaw/openclaw/issues/36182 | [Bug]: Perplexity Search API error (401) with OpenRouter | ### Bug type
Regression (worked before, now fails)
### Summary
Using Perplexity search engine via OpenRouter it throws an error since 2026.3.3 update.
The API Key is working since it's the same used for the main model.
### Steps to reproduce
1. Configure Perplexity as search engine with OpenRouter
"tools": {
... | closed | not_planned | true | 16 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T13:24:39Z | 2026-04-25T04:18:19Z | 2026-04-22T04:33:25Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | andres-lindo | 71,335,326 | MDQ6VXNlcjcxMzM1MzI2 | User | false |
openclaw/openclaw | 4,028,114,257 | I_kwDOQb6kR87wGCVR | 36,157 | https://github.com/openclaw/openclaw/issues/36157 | https://api.github.com/repos/openclaw/openclaw/issues/36157 | [Bug]: 在我代理api到cloudfare之后，开始无法识别图片了，上传图片之后报错，并且无法继续使用openclaw了 | ### Bug type
Regression (worked before, now fails)
### Summary
报错如下400 Item 'fc_0c525c4a4ce4502f0069a8f7c07dc88190b6f50347f7cefd23' of type 'function_call' was provided without its required 'reasoning' item: 'rs_0c525c4a4ce4502f0069a8f7ad168881908544e83a9e64b642'.
<img width="2216" height="984" alt="Image" src="htt... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T13:06:07Z | 2026-04-25T04:18:20Z | 2026-04-22T04:33:24Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | Echosssy | 114,723,111 | U_kgDOBtaJJw | User | false |
openclaw/openclaw | 4,028,056,836 | I_kwDOQb6kR87wF0UE | 36,142 | https://github.com/openclaw/openclaw/issues/36142 | https://api.github.com/repos/openclaw/openclaw/issues/36142 | Silent empty response when LLM rate limit hits mid-turn (after tool calls) | ## Bug Description
When the LLM API returns a rate limit (429) or overload error **after tool calls have already completed** in a turn, the assistant response is completely empty/silent. No error message is shown to the user, no indication of what went wrong, and the conversation appears stuck.
## Steps to Reproduce
... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-05T12:55:31Z | 2026-04-25T04:18:21Z | 2026-04-22T04:33:22Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | JaewookYou | 23,391,385 | MDQ6VXNlcjIzMzkxMzg1 | User | false |
openclaw/openclaw | 4,028,031,364 | I_kwDOQb6kR87wFuGE | 36,134 | https://github.com/openclaw/openclaw/issues/36134 | https://api.github.com/repos/openclaw/openclaw/issues/36134 | [Bug]: sessions_spawn silently falls back to Anthropic (set fallback model) when targeting Google Gemini models on spawned subagent tasks. | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When sessions_spawn targets a Google/Gemini model via either alias ("gemini") or full provider ID ("google/gemini-3.1-flash-lite-preview"), the session returns modelApplied: true but the sub-agent actually runs on the configured fallback pr... | closed | not_planned | true | 4 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-05T12:50:33Z | 2026-04-25T04:18:22Z | 2026-04-22T04:33:20Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | Workz777 | 68,589,694 | MDQ6VXNlcjY4NTg5Njk0 | User | false |
openclaw/openclaw | 4,028,003,463 | I_kwDOQb6kR87wFnSH | 36,119 | https://github.com/openclaw/openclaw/issues/36119 | https://api.github.com/repos/openclaw/openclaw/issues/36119 | Feature: Mid-run profile rotation on 429 + configurable rate limit cooldown | ## Feature Request: Improve 429 rate limit handling with multi-profile auth rotation
### Problem
When running multiple agents (5 agents + 17 cron jobs) with multiple Anthropic auth profiles (`auth.order`), 429 rate limit errors don't trigger profile rotation **within the same run's retry loop**. The current behavior:... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T12:44:27Z | 2026-04-25T04:18:22Z | 2026-04-22T04:33:18Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | Kenneth77chu | 141,257,376 | U_kgDOCGtqoA | User | false |
openclaw/openclaw | 4,027,922,186 | I_kwDOQb6kR87wFTcK | 36,088 | https://github.com/openclaw/openclaw/issues/36088 | https://api.github.com/repos/openclaw/openclaw/issues/36088 | [Bug]: openclaw tui route error | ### Bug type
Regression (worked before, now fails)
### Summary
## Bug 描述
TUI 的 `--deliver` 选项无法正常工作：使用 `openclaw tui --deliver` 启动时，助手回复不会显示在 TUI 中，而是被路由到 webchat channel。
## 环境信息
- OpenClaw 版本：2026.3.2
- 操作系统：Linux
- 节点：Node.js v22.22.0
### Steps to reproduce
1. 运行 `openclaw tui --deliver --session main`
2. 发送一... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T12:27:44Z | 2026-04-25T04:18:23Z | 2026-04-22T04:33:16Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | hangsa | 10,755,520 | MDQ6VXNlcjEwNzU1NTIw | User | false |
openclaw/openclaw | 4,027,910,043 | I_kwDOQb6kR87wFQeb | 36,082 | https://github.com/openclaw/openclaw/issues/36082 | https://api.github.com/repos/openclaw/openclaw/issues/36082 | Execution Plan: main-agent autonomy/proactivity hardening (silent-stall zero-tolerance) | ## Objective\nExecute a concrete plan to maximize autonomous/proactive reliability of main-agent orchestration while keeping security-first constraints.\n\n## Scope\n- Never-silent completion guarantees\n- Spawn verification hardening completion\n- Operational event-class routing policy\n- Session hygiene defaults\n\n#... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T12:25:07Z | 2026-04-25T04:18:24Z | 2026-04-22T04:33:15Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | ArgadronRey | 183,376,875 | U_kgDOCu4b6w | User | false |
openclaw/openclaw | 4,027,824,670 | I_kwDOQb6kR87wE7oe | 36,066 | https://github.com/openclaw/openclaw/issues/36066 | https://api.github.com/repos/openclaw/openclaw/issues/36066 | Feature Request: Enable /reasoning and other commands via LINE OA Channel | Currently, /reasoning and other direct commands (e.g., /status, /tools health) only seem to affect the Main Session (Web Console) and do not persist or activate within Channel Sessions (like LINE OA).
This leads to user confusion and a degraded experience, as users expect commands to work where they are typing them.
... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T12:06:44Z | 2026-04-25T04:18:25Z | 2026-04-22T04:33:13Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | Piboonsak | 17,107,792 | MDQ6VXNlcjE3MTA3Nzky | User | false |
openclaw/openclaw | 4,027,801,168 | I_kwDOQb6kR87wE15Q | 36,059 | https://github.com/openclaw/openclaw/issues/36059 | https://api.github.com/repos/openclaw/openclaw/issues/36059 | BlueBubbles 每30分钟 stale-socket 重连问题 | ## 问题描述
每30分钟（heartbeat interval）health-monitor 检测到 bluebubbles WebSocket stale-socket 并自动重启。
日志：
## 环境
- OpenClaw: 2026.3.2
- BlueBubbles: Mac 版
- Gateway heartbeat: 30分钟
## 期望
1. 能否提供可配置的 health-check 间隔参数？
2. 或者根本性解决 BlueBubbles 假死问题
## 相关日志
- 问题出现在每次 heartbeat 时，BlueBubbles 本身运行正常，但 WebSocket 连接会被判定为 stale | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-05T12:02:13Z | 2026-04-22T04:30:31Z | 2026-04-19T21:44:34Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | 1960697431 | 72,795,243 | MDQ6VXNlcjcyNzk1MjQz | User | false |
openclaw/openclaw | 4,027,792,706 | I_kwDOQb6kR87wEz1C | 36,057 | https://github.com/openclaw/openclaw/issues/36057 | https://api.github.com/repos/openclaw/openclaw/issues/36057 | Bug: openai-responses custom providers can drop effective tool_choice on streamSimple path (no tool calls) | ## Bug type
Behavior bug (incorrect output/state without crash)
## Summary
For `api: openai-responses` on non-`openai` providers (e.g. custom `sub2api-gpt`), OpenClaw can send requests with `tools` present but `tool_choice` missing/null on the streamSimple HTTP path, leading to text-only completions and no tool call e... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-05T12:00:18Z | 2026-04-25T04:18:26Z | 2026-04-22T04:33:11Z | CONTRIBUTOR | null | 20260425T060021Z | 2026-04-25T06:00:21Z | CharZhou | 17,255,546 | MDQ6VXNlcjE3MjU1NTQ2 | User | false |
openclaw/openclaw | 4,027,786,514 | I_kwDOQb6kR87wEyUS | 36,056 | https://github.com/openclaw/openclaw/issues/36056 | https://api.github.com/repos/openclaw/openclaw/issues/36056 | [Bug]: origin not allowed (open the Control UI from the gateway host or allow it in gateway.controlUi.allowedOrigins) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Page prompt: origin not allowed (open the Control UI from the gateway host or allow it in gateway.controlUi.allowedOrigins)
页面提示：origin not allowed (open the Control UI from the gateway host or allow it in gateway.controlUi.allowedOrigins)... | closed | not_planned | true | 5 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-05T11:58:57Z | 2026-04-25T04:18:28Z | 2026-04-22T04:33:09Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | admin6a | 152,629,102 | U_kgDOCRjvbg | User | false |
openclaw/openclaw | 4,027,718,541 | I_kwDOQb6kR87wEhuN | 36,035 | https://github.com/openclaw/openclaw/issues/36035 | https://api.github.com/repos/openclaw/openclaw/issues/36035 | Memory SQLite should use WAL journal mode by default | ## Problem
Memory index SQLite database (`~/.openclaw/memory/main.sqlite`) is created with `journal_mode=delete` (SQLite default). This causes frequent database corruption when the gateway receives SIGTERM during write operations (restart, update, config changes).
## Impact
- Users see "database disk image is malforme... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T11:45:06Z | 2026-04-25T04:18:29Z | 2026-04-22T04:33:07Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | nguathen | 213,979,139 | U_kgDODMEQAw | User | false |
openclaw/openclaw | 4,027,712,336 | I_kwDOQb6kR87wEgNQ | 36,033 | https://github.com/openclaw/openclaw/issues/36033 | https://api.github.com/repos/openclaw/openclaw/issues/36033 | Feishu: block-only turns can drop replies (dispatch ends with replies=0) | ## Problem
Feishu DM inbound can be processed successfully by the gateway/agent, but no reply is delivered to the Feishu client when the turn emits only `block` payloads and no `final` payload.
Symptoms observed in production:
- Inbound DM is logged and routed.
- Agent session contains assistant text.
- Feishu dispa... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T11:43:50Z | 2026-04-25T04:18:30Z | 2026-04-22T04:33:05Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | godisawake | 26,685,911 | MDQ6VXNlcjI2Njg1OTEx | User | false |
openclaw/openclaw | 4,027,666,835 | I_kwDOQb6kR87wEVGT | 36,017 | https://github.com/openclaw/openclaw/issues/36017 | https://api.github.com/repos/openclaw/openclaw/issues/36017 | Health-monitor falsely detects Discord connections as 'stuck' during LLM inference, causing restart thrash and 'Gateway is draining' errors | ## Description
Health-monitor detects Discord channel connections as "stuck" and repeatedly restarts them, causing `Gateway is draining` errors for agents attempting to send messages during the restart window. This creates a cycle where connections are restarted every ~10 minutes.
## Environment
- **OpenClaw version... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-05T11:34:02Z | 2026-04-25T04:18:30Z | 2026-04-22T04:33:03Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | Pygmalione | 214,871,471 | U_kgDODM6trw | User | false |
openclaw/openclaw | 4,027,658,543 | I_kwDOQb6kR87wETEv | 36,013 | https://github.com/openclaw/openclaw/issues/36013 | https://api.github.com/repos/openclaw/openclaw/issues/36013 | [Model Behavior] Gemini 3.1 Pro: Instruction misinterpretation + repetitive output loops | ## Summary
When using gemini-3.1-pro-preview via OpenRouter in OpenClaw, the model exhibited two critical issues:
1. Instruction misinterpretation: Failed to understand delegation context
2. Repetitive output loop: Generated thousands of repeated sentences
## Environment
- Model: openrouter/google/gemini-3.1-pro-previ... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T11:32:04Z | 2026-04-25T04:18:31Z | 2026-04-22T04:32:59Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | vichuys | 72,035,014 | MDQ6VXNlcjcyMDM1MDE0 | User | false |
openclaw/openclaw | 4,027,650,528 | I_kwDOQb6kR87wERHg | 36,010 | https://github.com/openclaw/openclaw/issues/36010 | https://api.github.com/repos/openclaw/openclaw/issues/36010 | Telegram media group fetch intermittently fails on Node.js 25 (undici IPv6) | ## Description
Intermittent `TypeError: fetch failed` when downloading photos from Telegram media groups (multiple photos sent at once). Single photo messages work fine.
## Environment
- **OpenClaw**: 2026.3.1 (stable)
- **Node.js**: 25.6.1 (arm64, macOS 26.3)
- **OS**: macOS Sequoia 26.3 (Apple Silicon)
## Error
``... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-05T11:30:11Z | 2026-04-25T04:18:32Z | 2026-04-22T04:32:58Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | VashikMart | 261,251,777 | U_kgDOD5JiwQ | User | false |
openclaw/openclaw | 4,027,537,286 | I_kwDOQb6kR87wD1eG | 35,975 | https://github.com/openclaw/openclaw/issues/35975 | https://api.github.com/repos/openclaw/openclaw/issues/35975 | [Bug]: OpenClaw fails with "上游返回了 N 个事件但未产生有效输出" when using third-party API relay with thinking mode | ### Bug type
Regression (worked before, now fails)
### Summary
**Description**
When using a third-party Anthropic API relay (proxy) that returns responses with extended thinking enabled, OpenClaw fails to parse the response and shows:
LLM error api_error: 上游返回了 19 个事件但未产生有效输出，请重试
Simple queries (without tools/syste... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T11:06:20Z | 2026-04-25T04:18:33Z | 2026-04-22T04:32:49Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | zhoumo-creator | 113,785,378 | U_kgDOBsg6Ig | User | false |
openclaw/openclaw | 4,027,489,160 | I_kwDOQb6kR87wDpuI | 35,965 | https://github.com/openclaw/openclaw/issues/35965 | https://api.github.com/repos/openclaw/openclaw/issues/35965 | Intermittent 'Invalid diff: now finding less tool calls!' on Max/Jason paths (after cron model override migration) | # Incident Record: Intermittent `Invalid diff: now finding less tool calls!` (Max/Jason)
- Date: 2026-03-05 (Asia/Taipei)
- Reporter: Marcus
- Impact: Intermittent task failures in Max/Jason paths; noisy/error outputs despite successful runs before/after.
## Symptom
- Error string:
- `Invalid diff: now finding less... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T10:56:17Z | 2026-04-25T04:18:33Z | 2026-04-22T04:32:45Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | marcustseng-agent | 262,137,098 | U_kgDOD5_lCg | User | false |
openclaw/openclaw | 4,027,454,954 | I_kwDOQb6kR87wDhXq | 35,959 | https://github.com/openclaw/openclaw/issues/35959 | https://api.github.com/repos/openclaw/openclaw/issues/35959 | Telegram provider stale-socket restart causes stuck message replay loop | ## Summary
Long-running agent sessions cause the Telegram provider to enter a stale-socket → restart → re-deliver loop, sending the same message to the user every 10 minutes indefinitely.
## Environment
- OpenClaw `2026.3.2`
- macOS (Darwin 24.6.0, Apple Silicon)
- Node v22.22.0
- Telegram channel enabled, `dmPolicy: ... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T10:50:15Z | 2026-04-25T04:18:34Z | 2026-04-22T04:32:44Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | stormchaser-creator | 260,266,770 | U_kgDOD4NbEg | User | false |
openclaw/openclaw | 4,027,430,027 | I_kwDOQb6kR87wDbSL | 35,957 | https://github.com/openclaw/openclaw/issues/35957 | https://api.github.com/repos/openclaw/openclaw/issues/35957 | Config migration issue: Invalid keys from v2026.3.2 cause gateway startup failure in v2026.3.3 | ## Description
Gateway fails to start after upgrading from v2026.3.2 to v2026.3.3 due to config schema mismatch. Keys that were valid in v2026.3.2 (`autoApproveDevices`, `requirePairing`) are rejected as invalid in v2026.3.3, causing 18+ hours of downtime.
## Environment
- **OpenClaw versions**: 2026.3.2 → 2026.3.3
... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T10:45:31Z | 2026-04-25T04:18:35Z | 2026-04-22T04:32:42Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | antryu | 8,435,340 | MDQ6VXNlcjg0MzUzNDA= | User | false |
openclaw/openclaw | 4,027,416,957 | I_kwDOQb6kR87wDYF9 | 35,954 | https://github.com/openclaw/openclaw/issues/35954 | https://api.github.com/repos/openclaw/openclaw/issues/35954 | [Community] DeepSeek Official Provider - Save 30% on API Costs | ## 🚀 New Provider Announcement
I have created a **DeepSeek Official Provider** for OpenClaw that offers significant cost savings and better performance compared to using HuggingFace Inference.
### 💰 Cost Comparison (per 1M tokens)
| Model | HuggingFace | **DeepSeek Official** | **Savings** |
|-------|-------------... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T10:43:18Z | 2026-04-25T04:18:36Z | 2026-04-22T04:32:40Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | BIGBEAR-cjx | 187,005,831 | U_kgDOCyV7hw | User | false |
openclaw/openclaw | 4,027,309,765 | I_kwDOQb6kR87wC97F | 35,929 | https://github.com/openclaw/openclaw/issues/35929 | https://api.github.com/repos/openclaw/openclaw/issues/35929 | [Feature Request] Ephemeral Live Reasoning Status for Telegram | ### Summary
Implement an ephemeral status message for Telegram reasoning/thinking streaming to replace the current prepended/appended text blocks.
### Problem
Currently, with `reasoning: stream` enabled on Telegram, the reasoning blocks are prepended/appended to the message as it generates. This leaves a large "wall o... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T10:24:05Z | 2026-04-25T04:18:36Z | 2026-04-22T04:32:32Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | mauk1us | 13,590,929 | MDQ6VXNlcjEzNTkwOTI5 | User | false |
openclaw/openclaw | 4,027,189,637 | I_kwDOQb6kR87wCgmF | 35,886 | https://github.com/openclaw/openclaw/issues/35886 | https://api.github.com/repos/openclaw/openclaw/issues/35886 | [Bug]: [acpx] ACP child processes not cleaned up after session TTL expires — zombie process accumulation | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When using the acpx backend for ACP sessions, child processes spawned during
a session (e.g. claude-agent-acp, metabase-mcp, node, npm, sh) are
not terminated after the session's TTL expires. This causes unbounded process
accumulation over... | closed | not_planned | true | 3 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-05T10:01:40Z | 2026-04-25T04:18:37Z | 2026-04-22T04:32:20Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | RickyShiJs | 7,312,454 | MDQ6VXNlcjczMTI0NTQ= | User | false |
openclaw/openclaw | 4,027,168,421 | I_kwDOQb6kR87wCbal | 35,884 | https://github.com/openclaw/openclaw/issues/35884 | https://api.github.com/repos/openclaw/openclaw/issues/35884 | [Bug]: warn Config warnings:\n- plugins.entries.feishu: plugin feishu: duplicate plugin id detected | ### Bug type
Regression (worked before, now fails)
### Summary
09:50:02 warn Config warnings:\n- plugins.entries.feishu: plugin feishu: duplicate plugin id detected; later plugin may be overridden (/root/.openclaw/extensions/feishu/index.ts)
Config warnings:\n- plugins.entries.feishu: plugin feishu: duplicate plugin... | closed | not_planned | true | 5 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T09:57:36Z | 2026-04-25T04:18:38Z | 2026-04-22T04:32:18Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | worlse | 40,832,932 | MDQ6VXNlcjQwODMyOTMy | User | false |
openclaw/openclaw | 4,027,111,722 | I_kwDOQb6kR87wCNkq | 35,871 | https://github.com/openclaw/openclaw/issues/35871 | https://api.github.com/repos/openclaw/openclaw/issues/35871 | Gateway dashboard shows wrong version (2026.2.24) after upgrading to 2026.3.2 | ## Description
After upgrading OpenClaw to v2026.3.2 via `npm install -g openclaw@latest`, the gateway dashboard and `openclaw status` still report `app 2026.2.24`.
## Environment
- **OS:** Ubuntu 22.04 (linux 6.8.0-100-generic x64)
- **Node:** v25.5.0
- **Install method:** npm global (`npm install -g openclaw`)
- *... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T09:46:51Z | 2026-04-25T04:18:39Z | 2026-04-22T04:32:17Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | alirezarezvani | 5,697,919 | MDQ6VXNlcjU2OTc5MTk= | User | false |
openclaw/openclaw | 4,027,087,274 | I_kwDOQb6kR87wCHmq | 35,864 | https://github.com/openclaw/openclaw/issues/35864 | https://api.github.com/repos/openclaw/openclaw/issues/35864 | bug: openclaw doctor silently strips unknown config keys without warning | ## Bug Report
### Summary
`openclaw doctor` silently removes unrecognised keys from `openclaw.json` during its repair pass without prompting the user or logging a clear warning. For a tool managing production-level agent configs, silent data loss is dangerous — users may not notice that intentional (or future-compatib... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T09:42:49Z | 2026-04-25T04:18:40Z | 2026-04-22T04:32:15Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | asz-hub | 224,570,580 | U_kgDODWKs1A | User | false |
openclaw/openclaw | 4,027,081,636 | I_kwDOQb6kR87wCGOk | 35,861 | https://github.com/openclaw/openclaw/issues/35861 | https://api.github.com/repos/openclaw/openclaw/issues/35861 | ACP Runtime: sessions_spawn fails with exit code 5 (ACP_TURN_FAILED) since v2026.3.2 | ## Environment
- OpenClaw: 2026.3.2 (latest)
- acpx: 0.1.15 (plugin-local)
- macOS 12.7.6 (x64)
- Codex ACP: 0.9.5 (gpt-5.3-codex)
## Symptoms
Every `sessions_spawn(runtime="acp", agentId="codex")` fails with:
```
AcpRuntimeError: acpx exited with code 5
code=ACP_TURN_FAILED
```
Sessions are created and immediately m... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-05T09:41:48Z | 2026-04-25T04:18:40Z | 2026-04-22T04:32:13Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | Flakedict | 41,390,664 | MDQ6VXNlcjQxMzkwNjY0 | User | false |
openclaw/openclaw | 4,027,075,656 | I_kwDOQb6kR87wCExI | 35,860 | https://github.com/openclaw/openclaw/issues/35860 | https://api.github.com/repos/openclaw/openclaw/issues/35860 | [Bug] Duplicate/multiplying responses in WebChat after context accumulates. Really need help. | ### Description
After context accumulates in a session, the assistant's responses start duplicating — beginning with 2 copies and increasing progressively with each exchange.
### Environment
- OpenClaw: 2026.3.2
- OS: Windows 11 + WSL2 Ubuntu 24.04
### Steps to Reproduce
1. Start a new session in WebChat
2. Continue ... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-05T09:40:45Z | 2026-04-25T04:18:41Z | 2026-04-22T04:32:11Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | laruje2837-hub | 262,743,066 | U_kgDOD6kkGg | User | false |
openclaw/openclaw | 4,027,075,041 | I_kwDOQb6kR87wCEnh | 35,859 | https://github.com/openclaw/openclaw/issues/35859 | https://api.github.com/repos/openclaw/openclaw/issues/35859 | [Bug] WebChat messages being forwarded to Telegram channel. I need help... | ### Description
Messages sent from the WebChat (Control UI) are being delivered to the Telegram channel as well.
### Environment
- OpenClaw: 2026.3.2
- OS: Windows 11 + WSL2 Ubuntu 24.04
- Channels configured: Telegram + WebChat
### Steps to Reproduce
1. Have both Telegram and WebChat channels configured
2. Send a me... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-05T09:40:40Z | 2026-04-25T04:18:42Z | 2026-04-22T04:32:09Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | laruje2837-hub | 262,743,066 | U_kgDOD6kkGg | User | false |
openclaw/openclaw | 4,027,063,725 | I_kwDOQb6kR87wCB2t | 35,857 | https://github.com/openclaw/openclaw/issues/35857 | https://api.github.com/repos/openclaw/openclaw/issues/35857 | feat: add pdfModel to config schema (agents.defaults.pdfModel) | ## Feature Gap
### Context
OpenClaw v2026.3.2 ships a native `pdf` tool that reads PDFs via Anthropic's native PDF API. The docs state the tool registers when OpenClaw can resolve a PDF-capable model config via:
1. `agents.defaults.pdfModel`
2. Fallback to `agents.defaults.imageModel`
3. Fallback to best-effort provi... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T09:38:41Z | 2026-04-25T04:18:42Z | 2026-04-22T04:32:08Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | asz-hub | 224,570,580 | U_kgDODWKs1A | User | false |
openclaw/openclaw | 4,027,032,071 | I_kwDOQb6kR87wB6IH | 35,851 | https://github.com/openclaw/openclaw/issues/35851 | https://api.github.com/repos/openclaw/openclaw/issues/35851 | [Bug]: Browser Relay: options page shows ON, but normal tabs stay ! and flip to “Relay not running” after attach | ### Bug type
Crash (process/app exits or hangs)
### Summary
Browser Relay Chrome extension options page shows ON, but attaching on normal tabs fails (tooltip flips from “Relay connected, click to attach to this tab” to “Relay not running” and badge stays !).
### Steps to reproduce
Stop any existing OpenClaw proces... | closed | not_planned | true | 5 | [
"bug",
"stale",
"bug:crash"
] | [] | 2026-03-05T09:33:24Z | 2026-04-25T04:18:43Z | 2026-04-22T04:32:06Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | knight0940 | 142,649,913 | U_kgDOCICqOQ | User | false |
openclaw/openclaw | 4,027,010,058 | I_kwDOQb6kR87wB0wK | 35,844 | https://github.com/openclaw/openclaw/issues/35844 | https://api.github.com/repos/openclaw/openclaw/issues/35844 | [Bug]: meta-ads-manager skill installation and Maton Gateway authentication issues | ### Bug type
Regression (worked before, now fails)
### Summary
Hello OpenClaw team,
I am encountering a persistent issue with the meta-ads-manager skill in my OpenClaw workspace.
Problem 1: Skill Installation Issues
After installing meta-ads-manager (via npx skills add openclaw/meta-ads-manager --force), the skill... | closed | not_planned | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T09:29:37Z | 2026-04-25T04:18:44Z | 2026-04-22T04:32:04Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | tinmacglorytins | 265,691,859 | U_kgDOD9Yi0w | User | false |
openclaw/openclaw | 4,026,999,149 | I_kwDOQb6kR87wByFt | 35,842 | https://github.com/openclaw/openclaw/issues/35842 | https://api.github.com/repos/openclaw/openclaw/issues/35842 | fix(openrouter): enforce explicit Claude Sonnet default instead of auto-routing | ## Problem
OpenRouter onboarding defaults to 'openrouter/auto', allowing provider/model drift at runtime (often routing to Gemini).
## Root Cause
The OPENROUTER_DEFAULT_MODEL_REF constant was set to 'openrouter/auto', a routing directive that delegates model selection to OpenRouter's API at request time.
## Solution
... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-05T09:27:38Z | 2026-04-25T04:18:45Z | 2026-04-22T04:32:02Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | Piboonsak | 17,107,792 | MDQ6VXNlcjE3MTA3Nzky | User | false |
openclaw/openclaw | 4,026,988,057 | I_kwDOQb6kR87wBvYZ | 35,838 | https://github.com/openclaw/openclaw/issues/35838 | https://api.github.com/repos/openclaw/openclaw/issues/35838 | feat: Context Budget Awareness — Working Buffer for Long Sessions | ## Summary
Long-running sessions (especially coding sessions via backup proxy) can exhaust the context window, leading to:
- **Persona drift** — system prompt rules get squeezed out
- **Tool hallucination** — tool schema details are lost
- **Memory fragmentation** — earlier conversation content is dropped
Currently O... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T09:25:46Z | 2026-04-25T04:18:45Z | 2026-04-22T04:32:00Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | cattia-claw | 259,205,888 | U_kgDOD3MrAA | User | false |
openclaw/openclaw | 4,026,963,969 | I_kwDOQb6kR87wBpgB | 35,830 | https://github.com/openclaw/openclaw/issues/35830 | https://api.github.com/repos/openclaw/openclaw/issues/35830 | [cron] Announce delivery uses hallucinated recipient IDs instead of configured delivery.to | ---
title: "OpenClaw Cron Delivery Bug Analysis"
date: 2026-03-05
tags: [openclaw, bug, delivery, feishu, cron]
source: "本地服务器 OpenClaw 2026.3.2 实例排查"
---
# OpenClaw Cron Delivery: Agent 可覆盖配置中的收件人 ID
## 问题记录
### 现象
服务器上运行 OpenClaw 2026.3.2，配置了多个 cron 定时任务（每日技术简报、工作总结、EvoMap 任务等），delivery 配置中明确指定了飞书收件人... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T09:21:31Z | 2026-04-25T04:18:46Z | 2026-04-22T04:31:58Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | ShaunLeeCN | 195,417,774 | U_kgDOC6XWrg | User | false |
openclaw/openclaw | 4,026,428,943 | I_kwDOQb6kR87v_m4P | 35,682 | https://github.com/openclaw/openclaw/issues/35682 | https://api.github.com/repos/openclaw/openclaw/issues/35682 | Feature: Add session delete commands (rm/clear) without full reset | ### Problem
`openclaw sessions` can list sessions and `sessions cleanup` can prune by policy, but there is no direct way to intentionally delete sessions.
Current options are either:
1. `openclaw reset` (too destructive), or
2. manually deleting files under `~/.openclaw/agents/*/sessions/` (unsafe and unintuitive).
#... | closed | not_planned | true | 6 | [] | [] | 2026-03-05T07:33:56Z | 2026-04-29T04:38:06Z | 2026-04-26T10:02:22Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | j30231 | 44,216,653 | MDQ6VXNlcjQ0MjE2NjUz | User | false |
openclaw/openclaw | 4,026,413,904 | I_kwDOQb6kR87v_jNQ | 35,675 | https://github.com/openclaw/openclaw/issues/35675 | https://api.github.com/repos/openclaw/openclaw/issues/35675 | Feature: Auto-populate session label with chat/group name from messaging providers | When using OpenClaw with messaging providers like Feishu (Lark), each group chat creates a separate session. In the web UI sessions list, these sessions only show the session ID and an empty `label` field, making it hard to identify which session belongs to which group.
**Current behavior:**
- Sessions created from gr... | closed | completed | true | 5 | [] | [] | 2026-03-05T07:30:28Z | 2026-04-28T04:32:30Z | 2026-04-26T01:30:53Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | KIDD-4869 | 52,339,351 | MDQ6VXNlcjUyMzM5MzUx | User | false |
openclaw/openclaw | 4,021,260,996 | I_kwDOQb6kR87vr5LE | 34,311 | https://github.com/openclaw/openclaw/issues/34311 | https://api.github.com/repos/openclaw/openclaw/issues/34311 | [Bug]: Toll callauto existiert nicht wird aber dauernd aufgerufen | ### Bug type
Regression (worked before, now fails)
### Summary
FALSCH! Ich habe SOFORT WIEDER FALSCH GEMACHT! callauto existiert nicht. Der korrekte Name ist functions.exec. Ich bin offensichtlich nicht in der Lage, mich zu korregieren. Ich gebe auf.
Bei Installation des Skill qmd von clawhub
### Steps to reproduc... | closed | completed | true | 6 | [
"bug",
"regression"
] | [] | 2026-03-04T10:30:27Z | 2026-04-27T05:03:18Z | 2026-04-24T19:21:54Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mysap24 | 26,843,182 | MDQ6VXNlcjI2ODQzMTgy | User | false |
openclaw/openclaw | 4,019,887,569 | I_kwDOQb6kR87vmp3R | 33,880 | https://github.com/openclaw/openclaw/issues/33880 | https://api.github.com/repos/openclaw/openclaw/issues/33880 | [Bug]: WhatsApp group chat messages not spawning agent sessions (OpenClaw 2026.3.2) | ### Bug type
Regression (worked before, now fails)
### Summary
Direct WhatsApp messages work, but group messages fail to trigger session spawning. The plugin version changed from `channel-web-BrzOOpQZ.js `(working) to `channel-web-sl83aqDv.js (broken)`. Group chats stopped working around 22:30 EST on March 3rd, 2026... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T03:58:01Z | 2026-04-25T04:18:05Z | 2026-04-22T04:30:30Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | ayeung4 | 48,194,334 | MDQ6VXNlcjQ4MTk0MzM0 | User | false |
openclaw/openclaw | 4,017,576,243 | I_kwDOQb6kR87vd1kz | 33,251 | https://github.com/openclaw/openclaw/issues/33251 | https://api.github.com/repos/openclaw/openclaw/issues/33251 | Slack sandbox attachments reject ~/.openclaw/workspaces paths resolved from /workspace media | ## Summary
When OpenClaw runs in sandbox mode, Slack file sends using the documented `message send + media` flow fail for files created under `/workspace/...`.
The sandbox path is resolved to a host path under `~/.openclaw/workspaces/...`, but Slack local media validation rejects that path as not being under an allow... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T16:25:08Z | 2026-04-20T04:26:50Z | 2026-04-17T04:24:02Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | raw34 | 1,996,768 | MDQ6VXNlcjE5OTY3Njg= | User | false |
openclaw/openclaw | 4,017,537,075 | I_kwDOQb6kR87vdsAz | 33,242 | https://github.com/openclaw/openclaw/issues/33242 | https://api.github.com/repos/openclaw/openclaw/issues/33242 | Opus 4.6 plain-text thinking leaks to users (not caught by thinking-tag strip) | ## Description
Follow-up from #12631. The `<thinking>` tag stripping fix works correctly, but Opus 4.6 also **"thinks out loud" in plain assistant text** — no XML tags, no thinking blocks — which gets delivered directly to customers.
## Current Behavior
The model writes internal reasoning as regular text that bypass... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-03T16:17:20Z | 2026-04-20T04:26:51Z | 2026-04-17T04:24:00Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | Valinainformatica | 213,354,025 | U_kgDODLeGKQ | User | false |
openclaw/openclaw | 4,017,525,100 | I_kwDOQb6kR87vdpFs | 33,239 | https://github.com/openclaw/openclaw/issues/33239 | https://api.github.com/repos/openclaw/openclaw/issues/33239 | Duplicate plugin loading warning when installing Feishu plugin via official install script | ## Environment
- **OpenClaw Version**: 2026.3.2 (installed on March 3rd, 2026)
- **Installation Method**: Official install script (`curl -fsSL https://openclaw.ai/install.sh | bash`)
- **Platform**: Linux server
- **Plugin**: Feishu (installed via npm package option during setup)
## Description
After installing OpenCl... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-03T16:14:51Z | 2026-04-20T04:26:52Z | 2026-04-17T04:23:58Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | ssdiwu | 16,416,080 | MDQ6VXNlcjE2NDE2MDgw | User | false |
openclaw/openclaw | 4,014,867,541 | I_kwDOQb6kR87vTgRV | 32,620 | https://github.com/openclaw/openclaw/issues/32620 | https://api.github.com/repos/openclaw/openclaw/issues/32620 | gateway restart always times out (60s) on systems without lsof | ## Bug
`openclaw gateway restart` **always times out after 60 seconds** on any Linux system without `lsof` installed, even though the gateway is running and healthy.
**Severity: HIGH** — affects every restart on Arch Linux, Alpine, minimal Debian/Ubuntu, containers, and any distro that doesn't ship `lsof` by default.... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-03T05:58:28Z | 2026-04-17T04:23:58Z | 2026-04-14T04:23:09Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | riftzen-bit | 139,470,135 | U_kgDOCFAlNw | User | false |
openclaw/openclaw | 4,014,875,258 | I_kwDOQb6kR87vTiJ6 | 32,623 | https://github.com/openclaw/openclaw/issues/32623 | https://api.github.com/repos/openclaw/openclaw/issues/32623 | Usage dashboard shows "Avg Tokens / Msg: 0" for agents with subagent sessions | ## Bug Description
The Usage dashboard's "Avg Tokens / Msg" shows `0` for agents that primarily run as subagent sessions (e.g., `worker` agent), even though individual subagent sessions correctly display token counts.
## Root Cause
Data path mismatch between token totals and message counts in the dashboard aggregati... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T06:01:00Z | 2026-04-17T04:23:57Z | 2026-04-14T04:23:13Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | kdotndot | 69,782,047 | MDQ6VXNlcjY5NzgyMDQ3 | User | false |
openclaw/openclaw | 4,014,871,195 | I_kwDOQb6kR87vThKb | 32,622 | https://github.com/openclaw/openclaw/issues/32622 | https://api.github.com/repos/openclaw/openclaw/issues/32622 | [Bug]: Return Msg: <400> InternalError.Algo.DataInspectionFailed: Input text data may contain inappropriate content. | ### Bug type
Regression (worked before, now fails)
### Summary
After using with Feishu, I got "<400> InternalError.Algo.DataInspectionFailed: Input text data may contain
inappropriate content." and can not recover from it.
<img width="2834" height="402" alt="Image" src="https://github.com/user-attachments/assets/... | closed | not_planned | true | 4 | [
"bug",
"stale"
] | [] | 2026-03-03T05:59:40Z | 2026-04-17T04:23:57Z | 2026-04-14T04:23:11Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | linshengli | 19,223,302 | MDQ6VXNlcjE5MjIzMzAy | User | false |
openclaw/openclaw | 4,014,913,856 | I_kwDOQb6kR87vTrlA | 32,632 | https://github.com/openclaw/openclaw/issues/32632 | https://api.github.com/repos/openclaw/openclaw/issues/32632 | [Bug]: WhatsApp selfChatMode fires on all outbound DMs, not just self-chat — agent responded to third-party contacts | ## Summary
When `selfChatMode: true` is set on the WhatsApp channel, the agent incorrectly processes and responds to messages sent in any DM thread — not just the user's self-chat (note-to-self).
## Config
```json
"whatsapp": {
"enabled": true,
"dmPolicy": "allowlist",
"selfChatMode": true,
"allowFrom": ["+1... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T06:13:00Z | 2026-04-17T04:23:56Z | 2026-04-14T04:23:17Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | econosopher | 10,812,302 | MDQ6VXNlcjEwODEyMzAy | User | false |
openclaw/openclaw | 4,014,923,432 | I_kwDOQb6kR87vTt6o | 32,633 | https://github.com/openclaw/openclaw/issues/32633 | https://api.github.com/repos/openclaw/openclaw/issues/32633 | [Bug]: sessions.json transcriptPath persisted as null — causes 400 on restart | ### Bug type
Regression (worked before, now fails)
### Summary
itle
sessions.json transcriptPath persisted as null — causes 400 on restart
Body
Description
All sessions in sessions.json have transcriptPath: null after gateway restart. When a session with transcriptPath: null tries to resume, it sends a broken messa... | closed | not_planned | true | 2 | [
"bug",
"stale"
] | [] | 2026-03-03T06:15:50Z | 2026-04-17T04:23:55Z | 2026-04-14T04:23:19Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | yuhui9 | 212,480,242 | U_kgDODKow8g | User | false |
openclaw/openclaw | 4,017,513,848 | I_kwDOQb6kR87vdmV4 | 33,236 | https://github.com/openclaw/openclaw/issues/33236 | https://api.github.com/repos/openclaw/openclaw/issues/33236 | [Bug] Model selection in Monica applies globally to all agents | ## Bug
When switching models on one agent in Monica, it switches models for ALL agents. Should be per-agent.
## Steps
1. Select any agent in Monica
2. Change its model
3. Notice ALL agents change to that model
## Expected
Each agent maintains its own model selection.
## Environment
OpenClaw 2026.3.2, macOS | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T16:12:25Z | 2026-04-20T04:26:53Z | 2026-04-17T04:23:54Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | tardissono | 2,653,812 | MDQ6VXNlcjI2NTM4MTI= | User | false |
openclaw/openclaw | 4,014,932,077 | I_kwDOQb6kR87vTwBt | 32,639 | https://github.com/openclaw/openclaw/issues/32639 | https://api.github.com/repos/openclaw/openclaw/issues/32639 | [Bug] Slack: agent react action fails in DMs, missing messageId auto-inference, missing currentMessageTs | ## Summary
Three related issues prevent the Slack `react` action from working reliably for agent-initiated emoji reactions:
### A) `slackTo` uses `user:` format for DMs — causes `channel_not_found`
In the Slack action handler, the target is computed as:
```typescript
const slackTo = isDirectMessage ? `user:${messag... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T06:18:21Z | 2026-04-17T04:23:54Z | 2026-04-14T04:23:26Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | CommanderCrowCode | 72,845,369 | MDQ6VXNlcjcyODQ1MzY5 | User | false |
openclaw/openclaw | 4,014,928,277 | I_kwDOQb6kR87vTvGV | 32,637 | https://github.com/openclaw/openclaw/issues/32637 | https://api.github.com/repos/openclaw/openclaw/issues/32637 | [Feature]: Per-agent working directory (cwd) separate from workspace | ### Summary
Support a per-agent `cwd` (or `workdir`) config key, separate from `workspace`, so multiple agents can share the same workspace files while operating in different working directories.
### Problem to solve
When running multiple agents (e.g. one per group chat), each agent needs to work in a different code... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T06:17:15Z | 2026-04-17T04:23:54Z | 2026-04-14T04:23:23Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Adi1231234 | 15,230,429 | MDQ6VXNlcjE1MjMwNDI5 | User | false |
openclaw/openclaw | 4,017,503,230 | I_kwDOQb6kR87vdjv- | 33,229 | https://github.com/openclaw/openclaw/issues/33229 | https://api.github.com/repos/openclaw/openclaw/issues/33229 | [slack] /session is broken | /session is broken in slack
<img width="487" height="54" alt="image" src="https://github.com/user-attachments/assets/5bafd9e9-b5c6-4eef-b6bb-6f808131e06f" />
_Originally posted by @sourman in https://github.com/openclaw/openclaw/issues/32696#issuecomment-3992052319_
| closed | not_planned | true | 10 | [
"stale"
] | [] | 2026-03-03T16:10:02Z | 2026-04-20T04:26:53Z | 2026-04-17T04:23:53Z | CONTRIBUTOR | null | 20260420T060038Z | 2026-04-20T06:00:38Z | sourman | 3,827,766 | MDQ6VXNlcjM4Mjc3NjY= | User | false |
openclaw/openclaw | 4,014,991,350 | I_kwDOQb6kR87vT-f2 | 32,651 | https://github.com/openclaw/openclaw/issues/32651 | https://api.github.com/repos/openclaw/openclaw/issues/32651 | Discord WebSocket disconnects during slow AI message processing | ## Description
Discord WebSocket connections silently die when AI message processing exceeds ~30 seconds. The health-monitor detects the stuck state and restarts the connection, but this introduces several minutes of downtime.
This is the same issue reported in #10743, which was closed by the stale bot without any co... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T06:36:15Z | 2026-04-17T04:23:53Z | 2026-04-14T04:23:27Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | nova-oc-bot | 262,968,592 | U_kgDOD6yVEA | User | false |
openclaw/openclaw | 4,015,071,001 | I_kwDOQb6kR87vUR8Z | 32,665 | https://github.com/openclaw/openclaw/issues/32665 | https://api.github.com/repos/openclaw/openclaw/issues/32665 | Telegram partial streaming: ~15s delay before streamed message cleanup on Pi | ## Description
On Telegram with `streaming: partial`, the streamed preview message takes ~15 seconds to be deleted after the final message is sent. This creates a visible "duplicate message" effect where both messages coexist briefly before the preview vanishes.
**Additional finding:** Setting `streaming: full` compl... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T06:59:52Z | 2026-04-17T04:23:52Z | 2026-04-14T04:23:29Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | soymarketing | 816,126 | MDQ6VXNlcjgxNjEyNg== | User | false |
openclaw/openclaw | 4,015,127,660 | I_kwDOQb6kR87vUfxs | 32,682 | https://github.com/openclaw/openclaw/issues/32682 | https://api.github.com/repos/openclaw/openclaw/issues/32682 | [Bug]: [Bug] 2026.3.1: Compiled gateway bundle reorders handlePluginRequest after canvasHost, causing 405 for plugin HTTP endpoints (e.g. Google Chat) | ### Bug type
Regression (worked before, now fails)
### Summary
In OpenClaw 2026.3.1, the compiled gateway bundle reorders handlePluginRequest after canvasHost, causing 405 for plugin HTTP endpoints. The TypeScript source has the correct ordering, but the bundler reorders them in the compiled output.
### Steps to re... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T07:15:53Z | 2026-04-17T04:23:51Z | 2026-04-14T04:23:34Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | asdf618 | 132,339,698 | U_kgDOB-NX8g | User | false |
openclaw/openclaw | 4,015,093,334 | I_kwDOQb6kR87vUXZW | 32,678 | https://github.com/openclaw/openclaw/issues/32678 | https://api.github.com/repos/openclaw/openclaw/issues/32678 | [Bug]: Feishu plugin bindings configuration not working - all messages routed to main agent | ### Bug type
Regression (worked before, now fails)
### Summary
The Feishu plugin does not respect the bindings configuration in openclaw.json. Regardless of how bindings are configured, all Feishu group messages are always dispatched to the main agent.
### Steps to reproduce
1. Configure multiple agents with bindi... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T07:06:12Z | 2026-04-17T04:23:51Z | 2026-04-14T04:23:31Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | tolontinomaggiowh182-sketch | 265,208,242 | U_kgDOD87Bsg | User | false |
openclaw/openclaw | 4,017,461,627 | I_kwDOQb6kR87vdZl7 | 33,220 | https://github.com/openclaw/openclaw/issues/33220 | https://api.github.com/repos/openclaw/openclaw/issues/33220 | Telegram: Support real-time streaming responses (editMessageText) | ## Feature Request: Real-Time Streaming Responses for Telegram
### Context
Telegram just announced native support for bots streaming responses in real-time ([Durov announcement](https://x.com/durov/status/2028455440862830970)). Text appears word-by-word like ChatGPT instead of waiting for the full reply to complete.
... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T16:01:18Z | 2026-04-20T04:26:54Z | 2026-04-17T04:23:50Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | bullstern | 11,908,218 | MDQ6VXNlcjExOTA4MjE4 | User | false |
openclaw/openclaw | 4,015,144,266 | I_kwDOQb6kR87vUj1K | 32,689 | https://github.com/openclaw/openclaw/issues/32689 | https://api.github.com/repos/openclaw/openclaw/issues/32689 | [Bug]: gateway health drops live runtime state and doctor warns for disabled allowlist channels | ## Summary
I updated OpenClaw on my VPS to `2026.3.2` and saw two false reports around channel state. The gateway health snapshot drops live runtime fields, and the doctor empty-allowlist scan still evaluates channels/accounts that are explicitly disabled.
## Actual behavior
- `openclaw channels status --json` shows... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-03T07:20:18Z | 2026-04-17T04:23:50Z | 2026-04-14T04:23:35Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | dorukardahan | 35,905,596 | MDQ6VXNlcjM1OTA1NTk2 | User | false |
openclaw/openclaw | 4,017,458,007 | I_kwDOQb6kR87vdYtX | 33,219 | https://github.com/openclaw/openclaw/issues/33219 | https://api.github.com/repos/openclaw/openclaw/issues/33219 | [Bug] macOS app crashes (SIGABRT) when Voice Wake trigger word fires — Swift exclusivity violation in VoiceWakeOverlayController.present() | ### Bug: macOS app crashes (SIGABRT) when Voice Wake trigger word fires
**App version:** 2026.3.2 (2026030290)
**macOS:** 26.1 (25B78)
**Hardware:** Mac Mini M4 (Mac16,10), ARM-64
**Mic:** USB PnP Audio Device (Fifine K669)
**Language:** Russian (ru-RU)
### Steps to reproduce
1. Open OpenClaw.app, go to Settin... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-03T16:00:38Z | 2026-04-20T04:26:55Z | 2026-04-17T04:23:49Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | polisadov | 29,985,731 | MDQ6VXNlcjI5OTg1NzMx | User | false |
openclaw/openclaw | 4,015,201,017 | I_kwDOQb6kR87vUxr5 | 32,712 | https://github.com/openclaw/openclaw/issues/32712 | https://api.github.com/repos/openclaw/openclaw/issues/32712 | [Bug]: Bot cannot retrieve card message content when a Feishu card is quoted and @mentioned in a group | ### Bug type
Regression (worked before, now fails)
### Summary
The function for the bot to obtain Feishu card message content worked properly in version v2026.2.26,
but stopped working after upgrading to v2026.3.1. The bot can only get: [Replying to: "[Interactive Card]"]
### Steps to reproduce
1. Configure the F... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T07:36:00Z | 2026-04-17T04:23:49Z | 2026-04-14T04:23:39Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | zhouwenhao123456 | 73,104,813 | MDQ6VXNlcjczMTA0ODEz | User | false |
openclaw/openclaw | 4,015,191,477 | I_kwDOQb6kR87vUvW1 | 32,704 | https://github.com/openclaw/openclaw/issues/32704 | https://api.github.com/repos/openclaw/openclaw/issues/32704 | Feature: before_agent_start hook should support { skip: true } to prevent agent run | ## Problem
When building a group message monitoring plugin (e.g., risk signal detection), the plugin needs to:
1. Receive all group messages (requireMention: false)
2. Classify messages via external API (fire-and-forget)
3. **NOT** run the main agent for non-@bot messages
Currently, `before_agent_start` can only modi... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T07:33:09Z | 2026-04-17T04:23:49Z | 2026-04-14T04:23:37Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | madsharksnow | 263,834,569 | U_kgDOD7nLyQ | User | false |
openclaw/openclaw | 4,015,220,770 | I_kwDOQb6kR87vU2gi | 32,721 | https://github.com/openclaw/openclaw/issues/32721 | https://api.github.com/repos/openclaw/openclaw/issues/32721 | [Bug]: Gemini 3.0 Flash outputs bare 'thought' text marker causing response duplication on Discord | ## Summary
Gemini 3.0 Flash Preview outputs a bare `thought` text marker (no XML tags) that bypasses all existing thinking-tag sanitization, causing the **entire response to be duplicated** in Discord.
Related to #6328 (locked) and #5946 — the XML tag fix does not cover this bare-text variant.
## Steps to reproduce
... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T07:42:16Z | 2026-04-17T04:23:48Z | 2026-04-14T04:23:41Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | goandon | 81,921,406 | MDQ6VXNlcjgxOTIxNDA2 | User | false |
openclaw/openclaw | 4,017,311,829 | I_kwDOQb6kR87vc1BV | 33,180 | https://github.com/openclaw/openclaw/issues/33180 | https://api.github.com/repos/openclaw/openclaw/issues/33180 | Telegram sendMessageDraft streaming causes messages to appear as Pinned Messages in DMs | ## Description
When `channels.telegram.streaming` is set to `partial` (the new default in 2026.3.2), messages streamed via `sendMessageDraft` in direct/private chats appear as **Pinned Messages** in the Telegram client (iOS tested).
## Steps to Reproduce
1. Set `channels.telegram.streaming: "partial"` in `openclaw.j... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-03T15:30:59Z | 2026-04-20T04:26:56Z | 2026-04-17T04:23:47Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | teebs4140 | 160,072,466 | U_kgDOCYqDEg | User | false |
openclaw/openclaw | 4,015,280,158 | I_kwDOQb6kR87vVFAe | 32,741 | https://github.com/openclaw/openclaw/issues/32741 | https://api.github.com/repos/openclaw/openclaw/issues/32741 | [Bug]: Select the model "Custom Provider", with the same configuration. After multiple prompts such as "Verification failed: This operation was aborted", the prompt "Verification Successful" will be displayed. | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Select the model "Custom Provider", with the same configuration. After multiple prompts such as "Verification failed: This operation was aborted", the prompt "Verification Successful" will be displayed.
What is the cause of this problem? I... | closed | not_planned | true | 3 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T08:00:22Z | 2026-04-17T04:23:47Z | 2026-04-14T04:23:46Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | yjy1992 | 23,351,898 | MDQ6VXNlcjIzMzUxODk4 | User | false |
openclaw/openclaw | 4,015,277,938 | I_kwDOQb6kR87vVEdy | 32,739 | https://github.com/openclaw/openclaw/issues/32739 | https://api.github.com/repos/openclaw/openclaw/issues/32739 | Slack Socket Mode: Provider keeps restarting in infinite loop | ## Description
Slack provider with Socket Mode enabled keeps restarting every 5-40 seconds in an infinite loop. No error messages are logged.
## Environment
- OpenClaw version: 2026.3.2
- Node.js version: 25.7.0 (also tested on 22.22.0)
- OS: macOS 15.3.0 (arm64)
## Configuration
diff
{
channels: {
slack: {
... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T07:59:39Z | 2026-04-17T04:23:47Z | 2026-04-14T04:23:45Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | oddrap | 16,490,645 | MDQ6VXNlcjE2NDkwNjQ1 | User | false |
openclaw/openclaw | 4,015,287,930 | I_kwDOQb6kR87vVG56 | 32,743 | https://github.com/openclaw/openclaw/issues/32743 | https://api.github.com/repos/openclaw/openclaw/issues/32743 | Bug: Discord message chunks delivered in reverse order | ## Bug Report
**Describe the bug**
When a long reply is automatically split into multiple chunks (via `textChunkLimit` / `maxLinesPerMessage`), the chunks sometimes arrive in Discord in **reverse order**. For example, a two-part reply where part 1 is "好，测一下：" and part 2 is "测试结果..." shows up in Discord with part 2 fir... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T08:02:50Z | 2026-04-17T04:23:46Z | 2026-04-14T04:23:48Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | otcombo | 8,352,042 | MDQ6VXNlcjgzNTIwNDI= | User | false |
openclaw/openclaw | 4,017,291,282 | I_kwDOQb6kR87vcwAS | 33,174 | https://github.com/openclaw/openclaw/issues/33174 | https://api.github.com/repos/openclaw/openclaw/issues/33174 | Telegram inbound polling broken in 2026.3.2 multi-account setup | ## Environment
- OpenClaw: 2026.3.2 (downgraded to 2026.2.26 as workaround)
- macOS (LaunchAgent), 2 Telegram accounts (different tokens)
- Config: multi-account with dmPolicy pairing/allowlist, allowFrom set
## Symptoms
- Outbound (sendMessage) works fine
- Inbound messages from Telegram are silently dropped — zero l... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T15:26:49Z | 2026-04-20T04:26:56Z | 2026-04-17T04:23:45Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | neverything | 368,120 | MDQ6VXNlcjM2ODEyMA== | User | false |
openclaw/openclaw | 4,015,306,459 | I_kwDOQb6kR87vVLbb | 32,745 | https://github.com/openclaw/openclaw/issues/32745 | https://api.github.com/repos/openclaw/openclaw/issues/32745 | [Bug]: Temporal decay applies to almost no files — DATED_MEMORY_PATH_RE regex too strict | ## Bug Description
Temporal decay for memory search (`memorySearch.query.hybrid.temporalDecay`) is effectively non-functional for most real-world workspaces because `DATED_MEMORY_PATH_RE` in `src/memory/temporal-decay.ts` only matches the exact pattern `memory/YYYY-MM-DD.md`.
## The Regex
```javascript
const DATED_M... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-03T08:08:35Z | 2026-04-17T04:23:45Z | 2026-04-14T04:23:50Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | ThatsR4d | 152,051,272 | U_kgDOCRAeSA | User | false |
openclaw/openclaw | 4,015,322,625 | I_kwDOQb6kR87vVPYB | 32,753 | https://github.com/openclaw/openclaw/issues/32753 | https://api.github.com/repos/openclaw/openclaw/issues/32753 | [Bug]: /acp slash command ignores inline arg, shows menu anyway | ### Bug type
Regression (worked before, now fails)
### Summary
When typing /acp close (or any action inline), Discord sends the interaction with action = null because the field uses static choices — so OpenClaw always falls back to the button picker menu.
Root cause: buildDiscordCommandOptions registers the action ... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T08:13:19Z | 2026-04-17T04:23:44Z | 2026-04-14T04:23:54Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | Warp1nx | 85,808,172 | MDQ6VXNlcjg1ODA4MTcy | User | false |
openclaw/openclaw | 4,015,316,592 | I_kwDOQb6kR87vVN5w | 32,752 | https://github.com/openclaw/openclaw/issues/32752 | https://api.github.com/repos/openclaw/openclaw/issues/32752 | [Bug]: skills.entries.<key>.enabled: true does not override requires.bins check | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Setting `skills.entries.<key>.enabled: true` in config does not force-enable a skill when its `requires.bins` condition is unmet. This is asymmetric with `enabled: false`, which force-disables a skill regardless of whether its requirements ... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T08:11:36Z | 2026-04-17T04:23:44Z | 2026-04-14T04:23:52Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | carrotRakko | 24,588,751 | MDQ6VXNlcjI0NTg4NzUx | User | false |
openclaw/openclaw | 4,017,252,748 | I_kwDOQb6kR87vcmmM | 33,167 | https://github.com/openclaw/openclaw/issues/33167 | https://api.github.com/repos/openclaw/openclaw/issues/33167 | feat(cron): CRON_DONE signal and condition-based job self-deletion | ## Problem: Watch-Until-Done Crons Have No Exit
OpenClaw cron jobs have no mechanism for **condition-based self-termination**. An agent session spawned by a cron job is isolated — it can't reach back into the scheduler to say "my work is done, stop running me."
**Real-world example:** A cron runs hourly to check whet... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T15:19:22Z | 2026-04-20T04:26:57Z | 2026-04-17T04:23:43Z | CONTRIBUTOR | null | 20260420T060038Z | 2026-04-20T06:00:38Z | minupla | 42,547,246 | MDQ6VXNlcjQyNTQ3MjQ2 | User | false |
openclaw/openclaw | 4,015,438,692 | I_kwDOQb6kR87vVrtk | 32,799 | https://github.com/openclaw/openclaw/issues/32799 | https://api.github.com/repos/openclaw/openclaw/issues/32799 | Session file lock not released when holding process dies (stale lock causes agent failures) | ## Problem
When an OpenClaw process holds a `.jsonl.lock` file and dies unexpectedly (crash, OOM-kill, SIGKILL), the lock file is **not cleaned up**. Subsequent sessions attempting to acquire the same lock time out (default 10s) and fail entirely.
This creates a silent failure mode: the agent stops responding to all ... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T08:45:03Z | 2026-04-17T04:23:43Z | 2026-04-14T04:23:58Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | tvirilha | 255,598,525 | U_kgDODzwfvQ | User | false |
openclaw/openclaw | 4,015,473,773 | I_kwDOQb6kR87vV0Rt | 32,807 | https://github.com/openclaw/openclaw/issues/32807 | https://api.github.com/repos/openclaw/openclaw/issues/32807 | [Bug]: Browser relay/node connected, but agent sessions remain messaging-only (no browser tab-read/tools) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Browser relay and node are connected correctly, but agent chat sessions still don’t get browser/tab-read tools, so runtime stays messaging-only.
### Steps to reproduce
1. Start OpenClaw gateway on macOS and install/enable Chrome Browser R... | closed | not_planned | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T08:54:12Z | 2026-04-17T04:23:42Z | 2026-04-14T04:23:59Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | shortshowmedia-coder | 265,224,996 | U_kgDOD88DJA | User | false |
openclaw/openclaw | 4,017,252,187 | I_kwDOQb6kR87vcmdb | 33,166 | https://github.com/openclaw/openclaw/issues/33166 | https://api.github.com/repos/openclaw/openclaw/issues/33166 | [Bug]: has a error after run 'pnpm install' | ### Bug type
Crash (process/app exits or hangs)
### Summary
<img width="2132" height="734" alt="Image" src="https://github.com/user-attachments/assets/1f103938-9383-4dcd-a900-6a664518ee15" />
### Steps to reproduce
1.git clone https://github.com/openclaw/openclaw.git
2.cd openclaw
3.pnpm install
### Expected beh... | closed | not_planned | true | 2 | [
"bug",
"stale",
"bug:crash"
] | [] | 2026-03-03T15:19:14Z | 2026-04-20T04:26:58Z | 2026-04-17T04:23:41Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | fuchanghai | 33,984,497 | MDQ6VXNlcjMzOTg0NDk3 | User | false |
openclaw/openclaw | 4,015,492,613 | I_kwDOQb6kR87vV44F | 32,812 | https://github.com/openclaw/openclaw/issues/32812 | https://api.github.com/repos/openclaw/openclaw/issues/32812 | [Bug]: New install, when I tried to enable the tools in the dashboard for agents can not click save | ### Bug type
Regression (worked before, now fails)
### Summary
<img width="1436" height="1052" alt="Image" src="https://github.com/user-attachments/assets/188e5153-3d8f-43d7-83cc-bf4d871f0001" />
This also happens with enable all and disable all.
### Steps to reproduce
1. Open web UI
2. Go to agents
3. Click M... | closed | not_planned | true | 9 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T08:58:53Z | 2026-04-17T04:23:41Z | 2026-04-14T04:24:03Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | kenswain | 160,466 | MDQ6VXNlcjE2MDQ2Ng== | User | false |
openclaw/openclaw | 4,015,477,907 | I_kwDOQb6kR87vV1ST | 32,808 | https://github.com/openclaw/openclaw/issues/32808 | https://api.github.com/repos/openclaw/openclaw/issues/32808 | [Bug]: WSL2 add feishu channel failed | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
openclaw channels add
🦞 OpenClaw 2026.3.1 (2a8ac97) — Your task has been queued; your dignity has been deprecated.
07:59:12 [plugins] plugins.allow is empty; discovered non-bundled plugins may auto-load: feishu (/home/ubuntu/.openclaw/e... | closed | not_planned | true | 3 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T08:55:14Z | 2026-04-17T04:23:41Z | 2026-04-14T04:24:01Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | boil-wsb | 71,118,530 | MDQ6VXNlcjcxMTE4NTMw | User | false |
openclaw/openclaw | 4,015,494,694 | I_kwDOQb6kR87vV5Ym | 32,814 | https://github.com/openclaw/openclaw/issues/32814 | https://api.github.com/repos/openclaw/openclaw/issues/32814 | Bug: resolveBrowserBaseUrl function missing return statement causes browser tool timeout | # Bug Report: Browser Tool Communication Failure
## 🐛 Issue Summary
**Title:** `resolveBrowserBaseUrl` function missing return statement causes browser tool timeout
**Severity:** High - Browser tool completely non-functional for embedded agent
**Version:** OpenClaw 2026.2.26
---
## 🔍 Problem Description
When u... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-03T08:59:21Z | 2026-04-17T04:23:40Z | 2026-04-14T04:24:05Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | yty25 | 32,146,427 | MDQ6VXNlcjMyMTQ2NDI3 | User | false |
openclaw/openclaw | 4,017,252,079 | I_kwDOQb6kR87vcmbv | 33,165 | https://github.com/openclaw/openclaw/issues/33165 | https://api.github.com/repos/openclaw/openclaw/issues/33165 | Feature: Send user notification when all failover models are exhausted (FailoverError) | ## Problem
When all configured models hit rate limits (or other errors) during failover, the system silently drops the user's message with no feedback. The user experiences this as "I sent a message but got no reply," and has to resend to trigger a retry.
**Observed behavior:**
- Primary model (e.g. `gemini-3-pro-pre... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-03T15:19:13Z | 2026-04-20T04:26:59Z | 2026-04-17T04:23:39Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | seasonhuang | 1,913,836 | MDQ6VXNlcjE5MTM4MzY= | User | false |
openclaw/openclaw | 4,015,566,005 | I_kwDOQb6kR87vWKy1 | 32,833 | https://github.com/openclaw/openclaw/issues/32833 | https://api.github.com/repos/openclaw/openclaw/issues/32833 | [Bug]: exec plugin install fails on v2026.3.2 | ### Bug type
Regression (worked before, now fails)
### Summary
The exec plugin failed to install in v2026.3.2, rendering the command-line functionality unusable.
### Steps to reproduce
1. Run ：openclaw plugins install exec
2. restart gateway
### Expected behavior
The exec plugin has been successfully installed... | closed | not_planned | true | 10 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T09:16:54Z | 2026-04-17T04:23:39Z | 2026-04-14T04:24:07Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | YohtHill | 147,355,007 | U_kgDOCMh1fw | User | false |
openclaw/openclaw | 4,015,613,994 | I_kwDOQb6kR87vWWgq | 32,838 | https://github.com/openclaw/openclaw/issues/32838 | https://api.github.com/repos/openclaw/openclaw/issues/32838 | [Bug]: clawsec-advisory-guardian: missing advisory_scope.mjs in openclaw-managed hook | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
## Summary
The `clawsec-advisory-guardian` managed hook fails to load because
`lib/advisory_scope.mjs` is missing from the distribution. The file is
imported by `matching.ts` but was never shipped.
## Versions affected
- 2026.2.27 (confi... | closed | not_planned | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T09:27:32Z | 2026-04-17T04:23:38Z | 2026-04-14T04:24:11Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | mike-dataduality | 56,106,819 | MDQ6VXNlcjU2MTA2ODE5 | User | false |
openclaw/openclaw | 4,015,601,792 | I_kwDOQb6kR87vWTiA | 32,836 | https://github.com/openclaw/openclaw/issues/32836 | https://api.github.com/repos/openclaw/openclaw/issues/32836 | Desktop Commander — Unified Claude Desktop control via AppleScript | ## Summary
OpenClaw currently has fragmented desktop control: `chatgpt_image_generate` (DALL-E via AppleScript), `browser` tool (web), `canvas` (rendering), but NO unified way to trigger Claude Desktop actions (code editor, MCP tools, Projects) from agent sessions.
**Proposal:** Add a `claude_desktop` tool to OpenCla... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T09:24:52Z | 2026-04-17T04:23:38Z | 2026-04-14T04:24:09Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | domnumb | 132,837,649 | U_kgDOB-rxEQ | User | false |
openclaw/openclaw | 4,017,237,141 | I_kwDOQb6kR87vciyV | 33,158 | https://github.com/openclaw/openclaw/issues/33158 | https://api.github.com/repos/openclaw/openclaw/issues/33158 | Sub-agents not traced in Langfuse — all activity attributed to main agent only | ## Description
When sub-agents (builder, librarian, growth, reporter, architect) are spawned by the main agent, their LLM calls do not appear as separate traces in Langfuse. All observed activity is attributed to the main agent only.
## Version
OpenClaw 2026.3.1 (2a8ac97)
## Setup
Langfuse keys configured as globa... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T15:16:27Z | 2026-04-20T04:26:59Z | 2026-04-17T04:23:37Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | patbig-afk | 257,161,533 | U_kgDOD1P5PQ | User | false |
openclaw/openclaw | 4,015,623,565 | I_kwDOQb6kR87vWY2N | 32,841 | https://github.com/openclaw/openclaw/issues/32841 | https://api.github.com/repos/openclaw/openclaw/issues/32841 | [Bug]: Telegram inbound/reply pipeline instability on OpenClaw (VPS, systemd user service) | ### Bug type
Regression (worked before, now fails)
### Summary
OpenClaw intermittently fails to reliably process Telegram inbound messages and/or send replies when running on a VPS as a systemd --user service. The issue appears as unstable delivery/processing behavior rather than a full permanent outage.
### Steps ... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-03T09:29:48Z | 2026-04-17T04:23:37Z | 2026-04-14T04:24:13Z | NONE | null | 20260417T163222Z | 2026-04-17T16:32:22Z | DenisGradel | 199,984,697 | U_kgDOC-uGOQ | User | false |
openclaw/openclaw | 4,017,228,359 | I_kwDOQb6kR87vcgpH | 33,157 | https://github.com/openclaw/openclaw/issues/33157 | https://api.github.com/repos/openclaw/openclaw/issues/33157 | [Bug]: Web UI redirects to /chat with "token missing" despite valid token – backend WebSocket works | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
After a clean installation and onboarding, visiting http://127.0.0.1:18789/ immediately redirects to /chat?session=main and shows unauthorized: gateway token missing. The gateway logs confirm token_missing. However, manual WebSocket test wi... | closed | not_planned | true | 4 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-03T15:14:52Z | 2026-04-20T04:27:00Z | 2026-04-17T04:23:36Z | NONE | null | 20260420T060038Z | 2026-04-20T06:00:38Z | liubrook | 28,746,227 | MDQ6VXNlcjI4NzQ2MjI3 | User | false |
openclaw/openclaw | 4,015,679,062 | I_kwDOQb6kR87vWmZW | 32,850 | https://github.com/openclaw/openclaw/issues/32850 | https://api.github.com/repos/openclaw/openclaw/issues/32850 | fix(telegram): missing setStatus integration prevents health monitoring | ## Summary
The Telegram provider lacks `setStatus` integration, causing `lastEventAt` and `lastInboundAt` timestamps to remain `null` even when messages are actively received. This prevents the channel health monitoring system from distinguishing between a functional quiet channel and a half-dead connection.
## Curre... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-03T09:41:48Z | 2026-04-17T04:23:36Z | 2026-04-14T04:24:14Z | CONTRIBUTOR | null | 20260417T163222Z | 2026-04-17T16:32:22Z | garnetlyx | 12,513,503 | MDQ6VXNlcjEyNTEzNTAz | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.