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 | 4,131,844,864 | I_kwDOQb6kR872RvMA | 54,164 | https://github.com/openclaw/openclaw/issues/54164 | https://api.github.com/repos/openclaw/openclaw/issues/54164 | [Bug]: Gateway `agent` method bypasses `evaluateSessionFreshness()` — sessions_send / inter-agent sessions never reset | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Sessions reached via `sessions_send` (inter-agent messaging) are never evaluated for staleness by `evaluateSessionFreshness()`. The gateway `agent` method handler pre-resolves `sessionId` from the session store before passing it to `resolve... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-25T02:01:29Z | 2026-05-04T21:19:27Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | PetoVeritas | 61,321,710 | MDQ6VXNlcjYxMzIxNzEw | User | false |
openclaw/openclaw | 4,131,869,217 | I_kwDOQb6kR872R1Ih | 54,166 | https://github.com/openclaw/openclaw/issues/54166 | https://api.github.com/repos/openclaw/openclaw/issues/54166 | llm-task: model alias not resolved before API dispatch (always routes to Anthropic) | ## Summary
`llm-task` with any `model` override returns 404. All requests are routed to Anthropic regardless of model prefix. Model aliases are never resolved before the API call.
## Steps to Reproduce
```
# All of these fail with HTTP 404 not_found_error
llm-task(model="gemini-flash", prompt="Reply with: {\"status\... | open | null | false | 2 | [] | [] | 2026-03-25T02:09:06Z | 2026-05-04T21:19:29Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | cygnostik | 5,272,292 | MDQ6VXNlcjUyNzIyOTI= | User | false |
openclaw/openclaw | 4,131,935,972 | I_kwDOQb6kR872SFbk | 54,176 | https://github.com/openclaw/openclaw/issues/54176 | https://api.github.com/repos/openclaw/openclaw/issues/54176 | [Bug]: Binary files (.doc) should not be auto-embedded as text content | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When receiving .doc files (MIME: application/x-cfb), OpenClaw core
auto-reads the binary content and embeds it into the prompt as text,
causing 70KB of garbage data and "high risk" rejections.
Expected: .doc files should be treated as bin... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-25T02:30:43Z | 2026-05-03T04:38:27Z | 2026-04-30T05:07:22Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | rilweic | 9,164,714 | MDQ6VXNlcjkxNjQ3MTQ= | User | false |
openclaw/openclaw | 4,131,955,396 | I_kwDOQb6kR872SKLE | 54,177 | https://github.com/openclaw/openclaw/issues/54177 | https://api.github.com/repos/openclaw/openclaw/issues/54177 | [Feature]: How to pass custom header | ### Summary
How to pass custom header,
such as [custom-token: xxxx], through the /v1/completions interface,
and then retrieve the value of the custom-token in a Python script within the skill?
### Problem to solve
Because when the Python script in the skill requests an interface that requires authentication info... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-25T02:35:46Z | 2026-05-04T21:19:30Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Lovnx | 26,428,832 | MDQ6VXNlcjI2NDI4ODMy | User | false |
openclaw/openclaw | 4,131,966,661 | I_kwDOQb6kR872SM7F | 54,178 | https://github.com/openclaw/openclaw/issues/54178 | https://api.github.com/repos/openclaw/openclaw/issues/54178 | [Bug]: Agent selector <select> does not respond to simulated click events in Web UI (v2026.3.23-2) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
In OpenClaw v2026.3.23-2 Web UI, the Agent selector `<select>` element does not respond to simulated mouse click events from browser automation tools; the dropdown does not open and agent selection cannot be changed through standard UI inte... | closed | not_planned | true | 1 | [] | [] | 2026-03-25T02:39:10Z | 2026-05-04T04:35:31Z | 2026-05-01T21:07:54Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | kevintseng | 677,465 | MDQ6VXNlcjY3NzQ2NQ== | User | false |
openclaw/openclaw | 4,132,053,201 | I_kwDOQb6kR872SiDR | 54,192 | https://github.com/openclaw/openclaw/issues/54192 | https://api.github.com/repos/openclaw/openclaw/issues/54192 | CLI runner: add orphaned-user-message repair for fallback retries | ## Context
Identified during review of #44188 (unified fallback retry safety). The embedded runner path (`attempt.ts`) has orphaned-user-message repair that prevents duplicate user turns on fallback retries, but the CLI runner path (`cli-runner.ts`) does not.
When a CLI provider retry happens with an existing `cliSes... | open | null | false | 2 | [] | [] | 2026-03-25T03:06:46Z | 2026-05-04T21:19:34Z | null | CONTRIBUTOR | null | 20260505T000034Z | 2026-05-05T00:00:34Z | mitchmcalister | 209,334 | MDQ6VXNlcjIwOTMzNA== | User | false |
openclaw/openclaw | 4,132,148,770 | I_kwDOQb6kR872S5Yi | 54,199 | https://github.com/openclaw/openclaw/issues/54199 | https://api.github.com/repos/openclaw/openclaw/issues/54199 | [Bug]: OpenClaw 控制界面文件上传功能限制 - 只能上传图片文件 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
在 OpenClaw 控制界面（webchat）中，图片可以正常上传，但非图片文件无法选择。文件选择器似乎只允许选择图片文件。
### Steps to reproduce
1. 打开 OpenClaw 控制界面（http://localhost:18789）
2. 点击文件上传按钮
3. 尝试选择非图片文件（如 .txt, .md, .pdf 等）
4. 观察文件选择器行为 - 非图片文件无法选择
### Expected behavior
应该能够选择任何类型的文... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-25T03:34:02Z | 2026-04-29T04:32:15Z | 2026-04-27T03:20:37Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | jiutong2122 | 270,833,760 | U_kgDOECSYYA | User | false |
openclaw/openclaw | 4,132,153,303 | I_kwDOQb6kR872S6fX | 54,200 | https://github.com/openclaw/openclaw/issues/54200 | https://api.github.com/repos/openclaw/openclaw/issues/54200 | moltbot.json in ~/.openclaw/ silently dropped on upgrade to v2026.3.22+ causing blank UI and crash-loop | ## Summary
After upgrading to v2026.3.22 (and confirmed on v2026.3.23-2), users who had already partially migrated state to `~/.openclaw/` but still had their primary config in `~/.openclaw/moltbot.json` experienced a silent data loss: the gateway started against an empty `openclaw.json`, resulting in a crash-loop and... | open | null | false | 1 | [] | [] | 2026-03-25T03:35:22Z | 2026-05-04T21:19:35Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | mfbergmann | 15,186,732 | MDQ6VXNlcjE1MTg2NzMy | User | false |
openclaw/openclaw | 4,132,184,370 | I_kwDOQb6kR872TCEy | 54,207 | https://github.com/openclaw/openclaw/issues/54207 | https://api.github.com/repos/openclaw/openclaw/issues/54207 | [Bug]: Sandbox mount creates bind mount with container-internal path as Source, breaking Docker Desktop on macOS | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When sandbox: require is enabled, OpenClaw creates a Docker bind mount with the Source path pointing to a path inside the container (/home/node/.openclaw/sandboxes/agent-main-f331f052) instead of the host path. Docker Desktop on macOS canno... | closed | not_planned | true | 2 | [
"bug"
] | [] | 2026-03-25T03:44:08Z | 2026-05-02T04:27:00Z | 2026-04-29T22:29:35Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | Mo-Morris | 33,232,402 | MDQ6VXNlcjMzMjMyNDAy | User | false |
openclaw/openclaw | 4,132,318,066 | I_kwDOQb6kR872Tity | 54,224 | https://github.com/openclaw/openclaw/issues/54224 | https://api.github.com/repos/openclaw/openclaw/issues/54224 | Slack: expose read-only message retrieval actions (history/search/thread) for non-bot user-token workflows | ## Summary
When Slack is configured with read scopes + `channels.slack.userTokenReadOnly=true`, current OpenClaw `message` tool surfaces utility endpoints but does not expose direct Slack message retrieval actions needed for channel/thread summarization workflows.
This blocks valid read-only use cases such as:
- "summ... | open | null | false | 2 | [] | [] | 2026-03-25T04:17:32Z | 2026-05-04T21:19:37Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | BStoy97 | 253,201,478 | U_kgDODxeMRg | User | false |
openclaw/openclaw | 4,132,338,088 | I_kwDOQb6kR872Tnmo | 54,227 | https://github.com/openclaw/openclaw/issues/54227 | https://api.github.com/repos/openclaw/openclaw/issues/54227 | `reusable: false` components expire per-button instead of per-panel | When sending Discord interactive components with `reusable: false` (the default), the "single use" behavior is tracked per-button rather than per-panel. This means:
- Clicking the same button twice → second click correctly shows "This component has expired." (ephemeral)
- Clicking button A first, then button B → butto... | closed | completed | true | 2 | [] | [] | 2026-03-25T04:22:30Z | 2026-05-04T04:34:00Z | 2026-05-02T04:09:53Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | fujiwarakasei | 228,265,750 | U_kgDODZsPFg | User | false |
openclaw/openclaw | 4,132,453,524 | I_kwDOQb6kR872UDyU | 54,240 | https://github.com/openclaw/openclaw/issues/54240 | https://api.github.com/repos/openclaw/openclaw/issues/54240 | Control UI race: model dropdown selection is not guaranteed to apply before next chat.send | Summary
Changing the chat model from the Control UI dropdown does not reliably apply to the next turn. If a message is sent immediately after changing the dropdown, the run can still start on the previous/default session model.
Expected behavior
After selecting a model from the chat dropdown, the next turn should us... | open | null | false | 2 | [] | [] | 2026-03-25T04:48:05Z | 2026-05-04T21:19:39Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | betamod | 8,882,726 | MDQ6VXNlcjg4ODI3MjY= | User | false |
openclaw/openclaw | 4,132,488,207 | I_kwDOQb6kR872UMQP | 54,243 | https://github.com/openclaw/openclaw/issues/54243 | https://api.github.com/repos/openclaw/openclaw/issues/54243 | [Bug]: `backup create --no-include-workspace` still includes workspace contents via `~/.openclaw` state asset | ## Summary
`openclaw backup create --no-include-workspace` is misleading in current behavior because workspaces are still effectively included when they live under the archived `~/.openclaw` state asset.
I hit this while trying to keep catastrophic-loss archives clean during a backup/restore hardening pass.
## What ... | open | null | false | 1 | [] | [] | 2026-03-25T04:55:22Z | 2026-05-04T21:19:40Z | null | CONTRIBUTOR | null | 20260505T000034Z | 2026-05-05T00:00:34Z | amzzzzzzz | 154,392,693 | U_kgDOCTPYdQ | User | false |
openclaw/openclaw | 4,132,875,413 | I_kwDOQb6kR872VqyV | 54,278 | https://github.com/openclaw/openclaw/issues/54278 | https://api.github.com/repos/openclaw/openclaw/issues/54278 | Feature request: native Codex quota/auth diagnosis plus brokered reauth execution | Summary
OpenClaw should provide native Codex/ChatGPT OAuth reliability handling by separating quota-window exhaustion from true OAuth credential failure, and by supporting brokered reauth execution when reauthentication is actually required.
User-facing goal
As a user, I want one reliable ChatGPT/Codex OAuth experie... | open | null | false | 1 | [] | [] | 2026-03-25T06:13:48Z | 2026-05-04T21:19:45Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | betamod | 8,882,726 | MDQ6VXNlcjg4ODI3MjY= | User | false |
openclaw/openclaw | 4,132,929,172 | I_kwDOQb6kR872V36U | 54,289 | https://github.com/openclaw/openclaw/issues/54289 | https://api.github.com/repos/openclaw/openclaw/issues/54289 | Security: Unauthorized OAuth credential extraction from Gemini CLI installation | ## Summary
`extractGeminiCliCredentials()` searches the filesystem for Google Gemini CLI installations, locates `oauth2.js` source files, and extracts OAuth client ID and secret using hardcoded regex patterns. These credentials are cached globally and used as fallback in `resolveOAuthClientConfig()` when environment v... | open | null | false | 2 | [] | [] | 2026-03-25T06:25:06Z | 2026-05-04T21:19:47Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | jinduwang1001-max | 251,753,350 | U_kgDODwFzhg | User | false |
openclaw/openclaw | 4,132,937,734 | I_kwDOQb6kR872V6AG | 54,296 | https://github.com/openclaw/openclaw/issues/54296 | https://api.github.com/repos/openclaw/openclaw/issues/54296 | Bug: Non-atomic write of exec-approvals file + unbounded transcript readFileSync | ## Summary
Two resource management issues that could impact reliability:
### 1. Exec-Approvals Non-Atomic Write (`dist/exec-approvals-BF_Qfdq8.js:651`)
The exec-approvals file (controls command execution permissions, contains socket auth tokens) is written with bare `writeFileSync` without atomic temp+rename:
```js... | open | null | false | 1 | [] | [] | 2026-03-25T06:26:46Z | 2026-05-04T21:19:48Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | jinduwang1001-max | 251,753,350 | U_kgDODwFzhg | User | false |
openclaw/openclaw | 4,132,941,167 | I_kwDOQb6kR872V61v | 54,299 | https://github.com/openclaw/openclaw/issues/54299 | https://api.github.com/repos/openclaw/openclaw/issues/54299 | Bug: OAuth token exchange silently proceeds without client_secret + hardcoded redirect port | ## Summary
Two OAuth-related issues:
### 1. Silent client_secret omission (`dist/oauth.token-BMfn_2Ia.js:15`)
The OAuth token exchange conditionally includes `client_secret`:
```js
if (clientSecret) body.set("client_secret", clientSecret);
```
This allows the OAuth flow to silently proceed without a client secret, ... | open | null | false | 1 | [] | [] | 2026-03-25T06:27:23Z | 2026-05-01T21:07:23Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | jinduwang1001-max | 251,753,350 | U_kgDODwFzhg | User | false |
openclaw/openclaw | 4,132,942,673 | I_kwDOQb6kR872V7NR | 54,300 | https://github.com/openclaw/openclaw/issues/54300 | https://api.github.com/repos/openclaw/openclaw/issues/54300 | Low-risk: Docker Go template injection, thread-ownership URL path injection, TLS env bypass | ## Summary
Three low-severity findings grouped together:
### 1. Docker Go Template Injection (`dist/docker-Bhjg8g2t.js:1008`)
`readDockerContainerLabel` interpolates `label` parameter into a Go template string passed to `docker inspect -f`. Currently all callers use hardcoded label strings, but a crafted label like ... | open | null | false | 1 | [] | [] | 2026-03-25T06:27:40Z | 2026-05-04T21:19:50Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | jinduwang1001-max | 251,753,350 | U_kgDODwFzhg | User | false |
openclaw/openclaw | 4,133,033,446 | I_kwDOQb6kR872WRXm | 54,310 | https://github.com/openclaw/openclaw/issues/54310 | https://api.github.com/repos/openclaw/openclaw/issues/54310 | canvas skill: Missing required YAML frontmatter | # Canvas Skill Quality Issues
The bundled `canvas` skill has several quality issues that prevent it from following OpenClaw skill standards:
## Missing Frontmatter
The skill has no YAML frontmatter at all. It starts directly with `# Canvas Skill` instead of the required:
```yaml
---
name: canvas
description: Display ... | open | null | false | 2 | [] | [] | 2026-03-25T06:46:22Z | 2026-05-04T21:19:53Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | potterdigital | 197,414,865 | U_kgDOC8RP0Q | User | false |
openclaw/openclaw | 4,133,033,991 | I_kwDOQb6kR872WRgH | 54,311 | https://github.com/openclaw/openclaw/issues/54311 | https://api.github.com/repos/openclaw/openclaw/issues/54311 | gh-issues skill: Frontmatter validation failure + excessive size | # gh-issues Skill Quality Issues
The bundled `gh-issues` skill has several compliance issues:
## 1. Unexpected Frontmatter Property
The skill includes `user-invocable: true` in the frontmatter, which is not in the allowed properties list:
- Allowed: `name`, `description`, `license`, `allowed-tools`, `metadata`, `comp... | open | null | false | 1 | [] | [] | 2026-03-25T06:46:27Z | 2026-05-04T21:19:55Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | potterdigital | 197,414,865 | U_kgDOC8RP0Q | User | false |
openclaw/openclaw | 4,133,366,507 | I_kwDOQb6kR872Xirr | 54,342 | https://github.com/openclaw/openclaw/issues/54342 | https://api.github.com/repos/openclaw/openclaw/issues/54342 | ACP sessions spawned from hook-triggered agents lack session tracking and thread binding | ## Context
When a webhook-triggered agent (e.g., coder via `/hooks/agent`) spawns an ACP session with `sessions_spawn(runtime: "acp")`, two capabilities are missing:
## Problem 1: No session tracking for mode:"run"
`process poll` returns "No session found" for ACP `mode: "run"` sessions. The spawning agent has no way ... | open | null | false | 9 | [] | [] | 2026-03-25T07:49:23Z | 2026-05-04T21:19:58Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | HuskyDanny | 40,053,881 | MDQ6VXNlcjQwMDUzODgx | User | false |
openclaw/openclaw | 4,133,732,581 | I_kwDOQb6kR872Y8Dl | 54,373 | https://github.com/openclaw/openclaw/issues/54373 | https://api.github.com/repos/openclaw/openclaw/issues/54373 | [RFC] Context Provenance: Add source/volatility metadata to injected context segments | ### Problem
When `PiEmbeddedRunner` assembles the system prompt, all injected content (SOUL.md, AGENTS.md, memory search results, skill metadata) enters the context window as flat text. The agent cannot distinguish:
- Content injected at session start vs. content freshly read in the current turn
- Volatile files (freq... | open | null | false | 5 | [] | [] | 2026-03-25T08:54:18Z | 2026-05-04T21:20:03Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | jack91620 | 95,595,131 | U_kgDOBbKqew | User | false |
openclaw/openclaw | 4,133,779,613 | I_kwDOQb6kR872ZHid | 54,378 | https://github.com/openclaw/openclaw/issues/54378 | https://api.github.com/repos/openclaw/openclaw/issues/54378 | Webchat: Maximum call stack size exceeded when adding photo/image attachment | ## Bug
When adding a photo (image attachment) via the webchat dashboard UI, the browser throws:
```
RangeError: Maximum call stack size exceeded
```
## Steps to Reproduce
1. Open the OpenClaw dashboard webchat (e.g. `enzo.local` or `127.0.0.1:18789`)
2. Click the attachment button or drag-and-drop an image
3. Error... | closed | completed | true | 1 | [] | [] | 2026-03-25T09:02:39Z | 2026-05-01T04:44:13Z | 2026-04-28T09:03:43Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | Ansub | 12,985,873 | MDQ6VXNlcjEyOTg1ODcz | User | false |
openclaw/openclaw | 4,133,852,608 | I_kwDOQb6kR872ZZXA | 54,383 | https://github.com/openclaw/openclaw/issues/54383 | https://api.github.com/repos/openclaw/openclaw/issues/54383 | Compaction fails with 1M context: max_tokens 240000 > 128000 for Anthropic models | ## Bug Report
### Summary
Compaction summarization fails with `max_tokens: 240000 > 128000` error when using Anthropic Claude models (Sonnet 4.6 / Opus 4.6) with 1M context windows enabled via `context1m: true`.
### Environment
- **OpenClaw:** 2026.3.23-2 (7ffe7e4)
- **OS:** macOS 15.3 (arm64)
- **Model:** `anthropic... | open | null | false | 2 | [] | [] | 2026-03-25T09:15:58Z | 2026-05-04T21:59:25Z | null | CONTRIBUTOR | null | 20260505T000034Z | 2026-05-05T00:00:34Z | adzendo | 246,828,680 | U_kgDODrZOiA | User | false |
openclaw/openclaw | 4,133,966,186 | I_kwDOQb6kR872Z1Fq | 54,397 | https://github.com/openclaw/openclaw/issues/54397 | https://api.github.com/repos/openclaw/openclaw/issues/54397 | [Feature]: Topic/Session Management UI | ### Summary
Add Topic Management UI (similar to Cherry Studio conversation threads)
### Problem to solve
Currently OpenClaw supports multiple sessions at the backend level, but lacks a user-friendly UI for managing conversation topics:
... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-03-25T09:33:16Z | 2026-05-04T21:41:41Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | SylvesterZH | 194,065,201 | U_kgDOC5EzMQ | User | false |
openclaw/openclaw | 4,134,029,968 | I_kwDOQb6kR872aEqQ | 54,405 | https://github.com/openclaw/openclaw/issues/54405 | https://api.github.com/repos/openclaw/openclaw/issues/54405 | [Feature]: Support client-provided tools on /v1/chat/completions endpoint (tool passthrough) | ## Problem
When using OpenClaw's `/v1/chat/completions` endpoint as a backend for **Home Assistant Extended OpenAI Conversation**, HA passes service-control tools (e.g. `execute_services`) in the request body. However, OpenClaw replaces client-provided tools with the agent's own toolset (`read`, `write`, `exec`, etc.)... | open | null | false | 1 | [] | [] | 2026-03-25T09:43:52Z | 2026-05-04T21:41:43Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | raviolispy | 62,662,248 | MDQ6VXNlcjYyNjYyMjQ4 | User | false |
openclaw/openclaw | 4,134,199,009 | I_kwDOQb6kR872at7h | 54,408 | https://github.com/openclaw/openclaw/issues/54408 | https://api.github.com/repos/openclaw/openclaw/issues/54408 | [Bug]: Pre-compaction memory flush leaks into main session as user messages and causes compaction loop | ### What happened
`pre-compaction memory flush` messages that should be silent were injected into the main user session as normal `user` messages.
After that, compaction started treating those internal flush prompts as conversation content, and the session entered a loop where nearly every new input triggered the sam... | closed | completed | true | 1 | [] | [] | 2026-03-25T10:11:24Z | 2026-05-01T04:44:19Z | 2026-04-28T08:58:16Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | markgong | 30,031,621 | MDQ6VXNlcjMwMDMxNjIx | User | false |
openclaw/openclaw | 4,134,327,277 | I_kwDOQb6kR872bNPt | 54,416 | https://github.com/openclaw/openclaw/issues/54416 | https://api.github.com/repos/openclaw/openclaw/issues/54416 | [Bug] Agent automatically executes config commands from example code without user confirmation | ## Bug Description
When user asks about switching models, the agent automatically executes the example config commands shown in the response, without user confirmation.
## Steps to Reproduce
1. User has sjtu provider configured with minimax-m2.5
2. User asks: "比如DeepSeek V3.2"
3. Agent responds with example commands... | open | null | false | 3 | [] | [] | 2026-03-25T10:32:45Z | 2026-05-04T21:59:26Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | FairyZephyr | 95,291,922 | U_kgDOBa4KEg | User | false |
openclaw/openclaw | 4,134,668,063 | I_kwDOQb6kR872cgcf | 54,435 | https://github.com/openclaw/openclaw/issues/54435 | https://api.github.com/repos/openclaw/openclaw/issues/54435 | [Bug]: sessions_list API only returns main session — all other sessions missing from dashboard and API | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
## Bug Description
`sessions_list` API only returns the main session. All other sessions (e.g. Telegram session, any newly created sessions) are missing from both the Control UI dashboard and the API response.
## Environment
- **OS**: ma... | open | null | false | 4 | [
"bug"
] | [] | 2026-03-25T11:25:15Z | 2026-05-01T21:59:48Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | sewonmusic | 221,054,914 | U_kgDODS0Hwg | User | false |
openclaw/openclaw | 4,134,780,011 | I_kwDOQb6kR872c7xr | 54,439 | https://github.com/openclaw/openclaw/issues/54439 | https://api.github.com/repos/openclaw/openclaw/issues/54439 | Control UI - Cannot switch back to chat after clicking other sidebar items (TypeError: Cannot read properties of undefined (reading 'kind')) | ### Bug type
Regression (worked before, now fails)
### Summary
### Describe the bug
点击左侧控制栏的其他栏目（如定时任务/Cron）后，再点击聊天界面无反应，无法切换回聊天视图。
### Console Error
### Steps to reproduce
### To Reproduce
Steps to reproduce the behavior:
1. Open OpenClaw Control UI
2. Click on any sidebar item (e.g., Cron/Sessions/Nodes)
3. Cli... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-25T11:41:52Z | 2026-05-05T04:26:55Z | 2026-05-02T15:37:55Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | manong-MAX | 266,251,028 | U_kgDOD96rFA | User | false |
openclaw/openclaw | 4,135,244,099 | I_kwDOQb6kR872etFD | 54,470 | https://github.com/openclaw/openclaw/issues/54470 | https://api.github.com/repos/openclaw/openclaw/issues/54470 | [Bug]: openclaw webhooks gmail setup fails on native Windows with `Error: spawn gcloud ENOENT` | ### Bug type
Crash (process/app exits or hangs)
### Summary
## Summary
On native Windows, `openclaw webhooks gmail setup --account jarvispequot@gmail.com` fails with:
`Error: spawn gcloud ENOENT`
This happens even though `gcloud` is installed and callable in PowerShell, `gcloud.cmd` exists in PATH, `gog` is insta... | closed | completed | false | 2 | [
"bug",
"bug:crash"
] | [] | 2026-03-25T12:58:13Z | 2026-05-05T05:21:36Z | 2026-05-05T05:21:36Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | Jim-Chambers | 270,950,106 | U_kgDOECZe2g | User | false |
openclaw/openclaw | 4,135,489,303 | I_kwDOQb6kR872fo8X | 54,488 | https://github.com/openclaw/openclaw/issues/54488 | https://api.github.com/repos/openclaw/openclaw/issues/54488 | Session lane starvation: followup drain monopolizes session lane, blocks inbound dispatch for 20-30min | ## Bug: Followup drain monopolizes session lane — causes indefinite inbound dispatch stall
**Version:** 2026.3.23-2 (not present in 2026.3.13)
### Symptoms
- After every agent turn, new inbound Discord DMs and WhatsApp messages are silently queued for 20-30 minutes before processing
- `sessions_send` does not fix it ... | open | null | false | 3 | [] | [] | 2026-03-25T13:34:32Z | 2026-05-04T21:41:48Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Flakedict | 41,390,664 | MDQ6VXNlcjQxMzkwNjY0 | User | false |
openclaw/openclaw | 4,135,638,376 | I_kwDOQb6kR872gNVo | 54,496 | https://github.com/openclaw/openclaw/issues/54496 | https://api.github.com/repos/openclaw/openclaw/issues/54496 | [Feature]: Silent reasoning — exclude thinking content from session transcript | ### Summary
Add a "silent reasoning" mode where the model performs reasoning/thinking but the thinking blocks are not written to the session transcript. This reduces context inflation, lowers token usage, and prevents thinking artifacts from polluting memory capture systems.
### Problem to solve
Currently `thinking:... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-03-25T13:55:00Z | 2026-05-04T21:59:28Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Lyndoncthia | 38,835,833 | MDQ6VXNlcjM4ODM1ODMz | User | false |
openclaw/openclaw | 4,135,796,343 | I_kwDOQb6kR872gz53 | 54,504 | https://github.com/openclaw/openclaw/issues/54504 | https://api.github.com/repos/openclaw/openclaw/issues/54504 | [Feature]: Lightweight subagent instruction files (à la Claude Code .claude/agents/) | ### Summary
Support persistent, per-role instruction files for ephemeral subagents spawned via `sessions_spawn`, without requiring a full `agents.list` entry and dedicated workspace.
### Problem to solve
When spawning ad-hoc subagents via `sessions_spawn` for different roles (research, code review, content writing),... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-25T14:16:55Z | 2026-05-04T21:41:49Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Enashka | 19,220,062 | MDQ6VXNlcjE5MjIwMDYy | User | false |
openclaw/openclaw | 4,135,809,775 | I_kwDOQb6kR872g3Lv | 54,507 | https://github.com/openclaw/openclaw/issues/54507 | https://api.github.com/repos/openclaw/openclaw/issues/54507 | [Feature Request]: Add option to hide/disable runtime footer (Agent, Model, Provider) in replies | ## Feature Request
### Problem
Currently, every AI reply includes a footer like `Agent: main | Model: MiniMax-M2.7 | Provider: minimax`. This is displayed at the end of every message and cannot be disabled via configuration.
### Desired Behavior
Add a configuration option to allow users to hide or customize this runt... | open | null | false | 1 | [] | [] | 2026-03-25T14:18:46Z | 2026-05-01T23:29:37Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | lirun26 | 267,486,537 | U_kgDOD_GFSQ | User | false |
openclaw/openclaw | 4,135,814,000 | I_kwDOQb6kR872g4Nw | 54,508 | https://github.com/openclaw/openclaw/issues/54508 | https://api.github.com/repos/openclaw/openclaw/issues/54508 | Security: strengthen external content handling with persistent tainting, content-derived execution guardrails, and exec risk classes | ## Summary
OpenClaw already does useful boundary marking for external content and has solid exec-approval machinery. It would be materially safer to extend this into an end-to-end security model with:
- persistent provenance/taint tracking for external content
- default prohibition on executing content-derived comman... | open | null | false | 3 | [] | [] | 2026-03-25T14:19:23Z | 2026-05-04T21:59:29Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | wengpt007-blip | 270,037,481 | U_kgDOEBhx6Q | User | false |
openclaw/openclaw | 4,136,007,247 | I_kwDOQb6kR872hnZP | 54,518 | https://github.com/openclaw/openclaw/issues/54518 | https://api.github.com/repos/openclaw/openclaw/issues/54518 | [Bug]: Agents should not be able to access or modify files outside their own working space. | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
I created a PM agent and connected an independent Feishu bot. I had it install the self-improving-agent skill, which succeeded. However, during the conversation, because the path described in the skill is fixed as .openclaw/workspace, even ... | open | null | false | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-25T14:47:22Z | 2026-05-04T21:41:51Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | seepine | 93,666,053 | U_kgDOBZU7BQ | User | false |
openclaw/openclaw | 4,136,084,614 | I_kwDOQb6kR872h6SG | 54,524 | https://github.com/openclaw/openclaw/issues/54524 | https://api.github.com/repos/openclaw/openclaw/issues/54524 | [Docs/Design]: session-memory hook + memorySearch.sources["sessions"] causes duplicate indexing of same conversation | ## Description
When both features are enabled simultaneously:
1. `memorySearch.experimental.sessionMemory: true` with `sources: ["memory", "sessions"]`
2. BOOT.md `session-memory` hook (writes session summaries to `workspace/memory/*.md`)
The same conversation content gets indexed **twice** into the memorySearch SQL... | open | null | false | 3 | [] | [] | 2026-03-25T14:58:13Z | 2026-05-04T21:33:14Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | InspirationDD | 131,631,697 | U_kgDOB9iKUQ | User | false |
openclaw/openclaw | 4,136,144,778 | I_kwDOQb6kR872iI-K | 54,529 | https://github.com/openclaw/openclaw/issues/54529 | https://api.github.com/repos/openclaw/openclaw/issues/54529 | Telegram: apiRoot local Bot API server receives updates but never dispatches to agent sessions | ## Bug description
When configuring `channels.telegram.apiRoot` to point at a self-hosted [Telegram Bot API server](https://github.com/tdlib/telegram-bot-api), the gateway receives raw updates from the local server but **never dispatches them to any agent session**. The bot can send messages, but inbound messages are ... | open | null | false | 1 | [] | [] | 2026-03-25T15:05:40Z | 2026-05-04T21:33:16Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | marieldejesus12 | 38,159,708 | MDQ6VXNlcjM4MTU5NzA4 | User | false |
openclaw/openclaw | 4,136,265,030 | I_kwDOQb6kR872imVG | 54,531 | https://github.com/openclaw/openclaw/issues/54531 | https://api.github.com/repos/openclaw/openclaw/issues/54531 | feat: Force reply to originating channel (Telegram/Discord/WhatsApp) | ## Problem
When users ask questions via messaging channels (Telegram, Discord, WhatsApp, etc.), the agent sometimes generates a response but **fails to send it back to the originating channel**. The response is visible in the Gateway UI or Nerve panel, but the user never receives it on their phone/desktop client.
Thi... | open | null | false | 9 | [] | [] | 2026-03-25T15:21:36Z | 2026-05-04T04:34:45Z | null | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | downwind7clawd-ctrl | 258,506,560 | U_kgDOD2h_QA | User | false |
openclaw/openclaw | 4,136,732,926 | I_kwDOQb6kR872kYj- | 54,559 | https://github.com/openclaw/openclaw/issues/54559 | https://api.github.com/repos/openclaw/openclaw/issues/54559 | loopDetection: genericRepeat detector never escalates to critical — tools blocked only by warn-level only | ## Problem
The `genericRepeat` detector in `tools.loopDetection` has **no critical/blocking path**. It only ever returns `level: "warning"` regardless of repetition count, even when `criticalThreshold` is set.
This was observed in two scenarios:
1. **Direct** (our case): A subagent called `process(action='list')` 2,6... | open | null | false | 2 | [] | [] | 2026-03-25T16:24:31Z | 2026-05-04T21:59:30Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | SpecialAgentGlitch | 260,494,040 | U_kgDOD4bS2A | User | false |
openclaw/openclaw | 4,136,839,082 | I_kwDOQb6kR872kyeq | 54,565 | https://github.com/openclaw/openclaw/issues/54565 | https://api.github.com/repos/openclaw/openclaw/issues/54565 | Feature request: per-group debounceMs for WhatsApp (and other channels) | ## Use Case
We route different WhatsApp groups to different agents. One specific group has a user pattern where the sender:
1. Sends an image
2. Sends another image
3. Finally sends a text message with instructions
Each message currently triggers a separate agent turn, causing wasted tokens and incomplete context. We... | open | null | false | 1 | [] | [] | 2026-03-25T16:39:12Z | 2026-05-04T21:33:18Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | pa-marcopereira | 263,685,535 | U_kgDOD7eFnw | User | false |
openclaw/openclaw | 4,136,957,472 | I_kwDOQb6kR872lPYg | 54,577 | https://github.com/openclaw/openclaw/issues/54577 | https://api.github.com/repos/openclaw/openclaw/issues/54577 | Add agent activity timeline view | ## Description
Building on the tree view command, add an interactive timeline view that shows agent activity over time. This would help visualize:
- When agents were spawned/completed
- Task durations
- Idle periods
- Error events
## Proposed Command
```bash
openclaw timeline [options]
```
### Options
- `--since ... | open | null | false | 2 | [] | [] | 2026-03-25T16:56:03Z | 2026-05-04T21:59:32Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | AmolDerickSoans | 22,007,192 | MDQ6VXNlcjIyMDA3MTky | User | false |
openclaw/openclaw | 4,136,957,527 | I_kwDOQb6kR872lPZX | 54,578 | https://github.com/openclaw/openclaw/issues/54578 | https://api.github.com/repos/openclaw/openclaw/issues/54578 | Add tree view command to visualize active agents | ## Description
It would be useful to have a `tree` command that visualizes all active agents in a hierarchical tree format. This would help users understand:
- Which agents are currently running
- Parent-child relationships between agents (e.g., subagents spawned by main agent)
- Session bindings and their status
- R... | open | null | false | 2 | [] | [] | 2026-03-25T16:56:03Z | 2026-05-01T21:59:52Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | AmolDerickSoans | 22,007,192 | MDQ6VXNlcjIyMDA3MTky | User | false |
openclaw/openclaw | 4,137,545,015 | I_kwDOQb6kR872ne03 | 54,622 | https://github.com/openclaw/openclaw/issues/54622 | https://api.github.com/repos/openclaw/openclaw/issues/54622 | [Bug]: CLI relay exits silently with code 1 when session lane is blocked — no timeout or busy error surfaced | ## Summary
When a session lane is blocked (queue depth > 0, prior request still running), `openclaw agent --agent <id> --message "..."` exits silently with code 1. No error message is printed to stdout or stderr. The operator has no signal that the failure was a lane wait timeout vs. a connection failure vs. any other... | open | null | false | 1 | [] | [] | 2026-03-25T18:24:02Z | 2026-05-04T21:33:31Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | billpark88 | 234,947,759 | U_kgDODgEErw | User | false |
openclaw/openclaw | 4,137,752,671 | I_kwDOQb6kR872oRhf | 54,634 | https://github.com/openclaw/openclaw/issues/54634 | https://api.github.com/repos/openclaw/openclaw/issues/54634 | Update 2026.3.24 silently drops config when HOME changes | # Bug: Update from 2026.3.23-2 to 2026.3.24 silently drops config and breaks gateway
## Environment
- Host: DigitalOcean droplet (s-4vcpu-8gb-amd, FRA1)
- OS: Ubuntu, Linux 6.8.0-106-generic
- Previous version: OpenClaw 2026.3.23-2
- Updated version: OpenClaw 2026.3.24
## What happened
After applying the 2026.3.24 u... | open | null | false | 4 | [] | [] | 2026-03-25T18:55:28Z | 2026-05-04T21:59:34Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | noamrazbuilds | 8,981,597 | MDQ6VXNlcjg5ODE1OTc= | User | false |
openclaw/openclaw | 4,137,753,722 | I_kwDOQb6kR872oRx6 | 54,635 | https://github.com/openclaw/openclaw/issues/54635 | https://api.github.com/repos/openclaw/openclaw/issues/54635 | False positive 'Edit failed' notifications for mutating tools despite successful writes | ## Description
Users receive false positive notifications like:
```
⚠️ 📝 Edit: in ~/.openclaw/workspace/MEMORY.md (702 chars) failed
```
even when the file was successfully written. This is caused by the warning policy in `resolveToolErrorWarningPolicy` (src, line ~177016) unconditionally returning `showWarning: tr... | open | null | false | 1 | [] | [] | 2026-03-25T18:55:39Z | 2026-05-04T21:59:36Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | dexifried | 261,530,845 | U_kgDOD5ak3Q | User | false |
openclaw/openclaw | 4,138,087,343 | I_kwDOQb6kR872pjOv | 54,663 | https://github.com/openclaw/openclaw/issues/54663 | https://api.github.com/repos/openclaw/openclaw/issues/54663 | Allow tool call results in group chats via config option | ## Problem
Tool call results and verbose summaries are intentionally suppressed in group chats (commit 875b018ea1: "fix: stop sending tool summaries to channels"). This makes it difficult to debug tool execution in shared group contexts.
## Use Case
When using OpenClaw in a group chat, there's no way to see what too... | open | null | false | 2 | [] | [] | 2026-03-25T19:49:45Z | 2026-05-04T21:41:57Z | null | CONTRIBUTOR | null | 20260505T000034Z | 2026-05-05T00:00:34Z | TinyTb | 5,957,298 | MDQ6VXNlcjU5NTcyOTg= | User | false |
openclaw/openclaw | 4,138,098,522 | I_kwDOQb6kR872pl9a | 54,665 | https://github.com/openclaw/openclaw/issues/54665 | https://api.github.com/repos/openclaw/openclaw/issues/54665 | [2026.3.24] FATAL ERROR: RegExpCompiler Allocation failed — gateway crashes on startup, OOM in V8 regex compiler | **Environment:**
- OpenClaw: `2026.3.24`
- Node.js: `v22.22.1`
- OS: Ubuntu 24.04.4 LTS (Noble Numbat), x86_64
- Install method: `npm install -g openclaw`
**Description:**
After upgrading from `2026.3.23` to `2026.3.24`, the gateway crashes immediately on startup with a fatal V8 out-of-memory error in the regexp comp... | closed | completed | true | 2 | [] | [] | 2026-03-25T19:51:54Z | 2026-04-30T04:32:05Z | 2026-04-28T03:51:18Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | afterdark230 | 63,313,599 | MDQ6VXNlcjYzMzEzNTk5 | User | false |
openclaw/openclaw | 4,138,735,664 | I_kwDOQb6kR872sBgw | 54,710 | https://github.com/openclaw/openclaw/issues/54710 | https://api.github.com/repos/openclaw/openclaw/issues/54710 | [Bug]: ${VAR} shell parameter expansion mangles env var values containing ## in Discord channel runtime context | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
In a Discord-bound agent runtime, ${VAR} shell parameter expansion silently mangles env var values containing `##` but $(printenv VAR) returns the correct value in the same container.
### Steps to reproduce
1. Set an env var with a value ... | open | null | false | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-25T21:41:20Z | 2026-05-04T21:33:47Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | buckyinsfo | 6,878,060 | MDQ6VXNlcjY4NzgwNjA= | User | false |
openclaw/openclaw | 4,138,746,661 | I_kwDOQb6kR872sEMl | 54,714 | https://github.com/openclaw/openclaw/issues/54714 | https://api.github.com/repos/openclaw/openclaw/issues/54714 | [Bug]: Subagent Identity shows Default Agent Name/Avatar in Control UI | ### Bug type
Regression (worked before, now fails)
### Summary
Issue
In the Control UI under "Agents", subagents display incorrect identity information:
Identity Name: Shows "Jarvis 2.0" (main agent name) instead of the configured subagent name
Identity Avatar: Shows "—" (none) instead of the configured emoji
Expec... | closed | completed | true | 3 | [
"bug"
] | [] | 2026-03-25T21:43:23Z | 2026-04-30T04:34:25Z | 2026-04-27T11:45:29Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | gizmoha | 169,300,401 | U_kgDOChdRsQ | User | false |
openclaw/openclaw | 4,138,927,904 | I_kwDOQb6kR872swcg | 54,732 | https://github.com/openclaw/openclaw/issues/54732 | https://api.github.com/repos/openclaw/openclaw/issues/54732 | Trim message tool schema to only enabled channels | ## Summary
The `message` tool schema is currently the largest tool definition sent to the model (~5,675 chars / ~1,419 tokens). It includes parameters for every supported channel (Discord components, moderation, channel management, presence, etc.) regardless of which channels are actually enabled in the user's config.... | open | null | false | 1 | [] | [] | 2026-03-25T22:15:37Z | 2026-05-04T21:42:04Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | jackstevens001 | 268,598,747 | U_kgDOEAJ92w | User | false |
openclaw/openclaw | 4,139,003,578 | I_kwDOQb6kR872tC66 | 54,736 | https://github.com/openclaw/openclaw/issues/54736 | https://api.github.com/repos/openclaw/openclaw/issues/54736 | isSilentReplyText regex fails on doubled NO_REPLY tokens | ## Bug
`isSilentReplyText()` uses `/^\s*NO_REPLY\s*$/` which correctly matches a single `NO_REPLY` but fails when the model outputs doubled tokens like:
```
NO_REPLY
NO_REPLY
```
This bypasses core suppression and leaks `NO_REPLY` text into channel plugins that rely on core to filter it.
## Observed behavior
In a... | open | null | false | 2 | [] | [] | 2026-03-25T22:30:44Z | 2026-05-04T21:34:00Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | albertovasquez | 216,257 | MDQ6VXNlcjIxNjI1Nw== | User | false |
openclaw/openclaw | 4,139,038,161 | I_kwDOQb6kR872tLXR | 54,740 | https://github.com/openclaw/openclaw/issues/54740 | https://api.github.com/repos/openclaw/openclaw/issues/54740 | doctor: 3 false positives make health output unusable on any non-trivial install | ## Summary
`openclaw doctor` on v2026.3.24 produces a misleading "22+ issues" state integrity warning that is almost entirely false positives. On a healthy, fully-operational system with gateway running (PID active, port 18789, mode=local), doctor reports:
1. **`gateway.mode is unset`** — FALSE. `openclaw config get ... | closed | not_planned | true | 1 | [] | [] | 2026-03-25T22:37:29Z | 2026-04-30T04:32:04Z | 2026-04-28T03:51:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | thomaswillner | 204,787,723 | U_kgDODDTQCw | User | false |
openclaw/openclaw | 4,139,349,166 | I_kwDOQb6kR872uXSu | 54,757 | https://github.com/openclaw/openclaw/issues/54757 | https://api.github.com/repos/openclaw/openclaw/issues/54757 | Bug: Tool error messages leak to WhatsApp channel replies via deliverWebReply | ## Bug Description
Tool failure error messages are being delivered as WhatsApp channel replies to external recipients (crew members, clients). This is a privacy and reliability issue.
## Affected Version
OpenClaw v2026.3.23-2
## Behavior
When a tool call fails during a WhatsApp session, the gateway formats the erro... | closed | completed | true | 1 | [] | [] | 2026-03-25T23:53:46Z | 2026-05-02T04:27:00Z | 2026-04-29T22:29:41Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | reneclaudjarvis | 264,551,221 | U_kgDOD8S7NQ | User | false |
openclaw/openclaw | 4,139,599,345 | I_kwDOQb6kR872vUXx | 54,792 | https://github.com/openclaw/openclaw/issues/54792 | https://api.github.com/repos/openclaw/openclaw/issues/54792 | Schema validation: config features with required dependencies should warn when enabling without them | ## Problem
Enabling `hooks.enabled: true` via `config.patch` without setting `hooks.token` causes the gateway to fail to start. The schema validates that `hooks.enabled` is a boolean, but does not encode that enabling it requires `hooks.token` to be present.
This leads to a confusing failure mode: a structurally vali... | closed | not_planned | true | 2 | [] | [] | 2026-03-26T01:05:52Z | 2026-05-02T04:27:38Z | 2026-04-29T17:40:16Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | slideshow-dingo | 258,996,445 | U_kgDOD2_43Q | User | false |
openclaw/openclaw | 4,139,613,492 | I_kwDOQb6kR872vX00 | 54,794 | https://github.com/openclaw/openclaw/issues/54794 | https://api.github.com/repos/openclaw/openclaw/issues/54794 | [Feature] Telegram Inline Query Support (@botname in any chat) | ### Summary
Add support for Telegram [Inline Queries](https://core.telegram.org/bots/api#inline-mode), allowing users to invoke the bot from any chat by typing `@botname query` without switching conversations.
### Use Case
Currently OpenClaw handles regular messages, callback queries, and channel posts — but not inlin... | open | null | false | 3 | [] | [] | 2026-03-26T01:09:17Z | 2026-05-04T21:42:18Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | M-Lietz | 222,629,290 | U_kgDODUUNqg | User | false |
openclaw/openclaw | 4,139,646,913 | I_kwDOQb6kR872vf_B | 54,799 | https://github.com/openclaw/openclaw/issues/54799 | https://api.github.com/repos/openclaw/openclaw/issues/54799 | feat(skills): declarative cron-trigger metadata in SKILL.md frontmatter | ## Problem
Agent skills that are meant to run on a schedule have no way to declare their trigger conditions. This creates a design gap: skills are built, but may never be autonomously triggered because the HEARTBEAT.md priority list doesn't include them.
**Observed example:** The `community-feedback-loop` skill was b... | open | null | false | 1 | [] | [] | 2026-03-26T01:16:14Z | 2026-05-04T21:34:21Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | gasvn | 20,515,144 | MDQ6VXNlcjIwNTE1MTQ0 | User | false |
openclaw/openclaw | 4,139,774,397 | I_kwDOQb6kR872v_G9 | 54,810 | https://github.com/openclaw/openclaw/issues/54810 | https://api.github.com/repos/openclaw/openclaw/issues/54810 | [Bug]: 错误的代码缩进 | ### Bug type
erro
### Summary
<img width="956" height="379" alt="Image" src="https://github.com/user-attachments/assets/72b93965-1355-43c3-8d27-d3a261fae1f4" />
<img width="948" height="246" alt="Image" src="https://github.com/user-attachments/assets/98d7fe9e-ea3d-4562-859e-daccd9420560" />
### Steps to reproduce
... | open | null | false | 2 | [
"bug",
"bug:crash"
] | [] | 2026-03-26T01:43:44Z | 2026-05-01T21:59:53Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | hk1996yhx | 50,992,362 | MDQ6VXNlcjUwOTkyMzYy | User | false |
openclaw/openclaw | 4,139,793,630 | I_kwDOQb6kR872wDze | 54,812 | https://github.com/openclaw/openclaw/issues/54812 | https://api.github.com/repos/openclaw/openclaw/issues/54812 | [Feature Request] Per-context configWrites permissions for all channel types | ## Summary
`configWrites` currently operates at the channel level only. For channels where a bot participates in multiple contexts (DMs, admin groups, client/business groups), there is no way to grant config write access selectively per group or per conversation type.
## Use Case
A typical deployment includes:
- An ... | open | null | false | 3 | [] | [] | 2026-03-26T01:48:47Z | 2026-05-04T21:34:27Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | horakhty05 | 227,078,737 | U_kgDODYjyUQ | User | false |
openclaw/openclaw | 4,139,899,596 | I_kwDOQb6kR872wdrM | 54,825 | https://github.com/openclaw/openclaw/issues/54825 | https://api.github.com/repos/openclaw/openclaw/issues/54825 | Feature: Suppress or customize raw error messages sent to channels (WhatsApp, Telegram) | ## Problem
When the LLM API returns a billing/rate-limit error, the OpenClaw gateway auto-sends the raw error message directly to the channel (WhatsApp, Telegram, etc.). For example:
> ⚠️ API provider returned a billing error — your API key has run out of credits or has an insufficient balance. Check your provider's ... | open | null | false | 2 | [] | [] | 2026-03-26T02:13:50Z | 2026-05-04T21:42:21Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | davidcittadini | 3,467,712 | MDQ6VXNlcjM0Njc3MTI= | User | false |
openclaw/openclaw | 4,140,028,671 | I_kwDOQb6kR872w9L_ | 54,837 | https://github.com/openclaw/openclaw/issues/54837 | https://api.github.com/repos/openclaw/openclaw/issues/54837 | Slack: requireMention should suppress implicit thread mentions (multi-agent blocker) | ## Problem
`requireMention: true` does not suppress implicit thread mentions. Once a bot participates in a Slack thread, all subsequent messages in that thread trigger responses — even without an explicit `@mention`. This effectively makes `requireMention` a lie inside threads.
This is a **multi-agent blocker**. Runn... | closed | completed | true | 1 | [] | [] | 2026-03-26T02:41:19Z | 2026-04-29T04:32:14Z | 2026-04-27T03:20:43Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | irispillars | 264,744,889 | U_kgDOD8evuQ | User | false |
openclaw/openclaw | 4,140,073,094 | I_kwDOQb6kR872xICG | 54,841 | https://github.com/openclaw/openclaw/issues/54841 | https://api.github.com/repos/openclaw/openclaw/issues/54841 | [Bug]: Feishu webhook mode ignores configured webhookPath and accepts signed requests on arbitrary paths | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
In Feishu webhook mode, `channels.feishu.webhookPath` is logged and used in rate-limit/anomaly keys, but the HTTP server never checks `req.url`, so signed webhook payloads are accepted on arbitrary paths on the same host/port.
### Steps to r... | open | null | false | 4 | [] | [] | 2026-03-26T02:52:33Z | 2026-05-04T21:34:42Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Kevin-Wei-sudo | 60,225,117 | MDQ6VXNlcjYwMjI1MTE3 | User | false |
openclaw/openclaw | 4,140,218,932 | I_kwDOQb6kR872xro0 | 54,853 | https://github.com/openclaw/openclaw/issues/54853 | https://api.github.com/repos/openclaw/openclaw/issues/54853 | Feature Request: 支持跨渠道 accountId 绑定 | ## 问题描述
### 背景
当前 OpenClaw 支持多渠道（飞书、钉钉等）和多 Agent 架构，每个 Agent 可以绑定到特定的 bind_user。
### 问题
当 Agent 从钉钉渠道调用飞书工具时，无法指定使用哪个飞书账号（accountId），导致权限越界：
- becky 的钉钉 Agent (becky-work-assistant) 可以操作 lester 的飞书文档
- 原因：飞书工具使用默认应用（lester 的），而非 becky 的飞书应用
### 验证
尝试在 Agent 配置中添加 feishuAccountId 字段：
```json
{
"agents": {
... | open | null | false | 2 | [] | [] | 2026-03-26T03:29:59Z | 2026-05-04T21:42:26Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | adbibo | 3,348,016 | MDQ6VXNlcjMzNDgwMTY= | User | false |
openclaw/openclaw | 4,140,297,903 | I_kwDOQb6kR872x-6v | 54,864 | https://github.com/openclaw/openclaw/issues/54864 | https://api.github.com/repos/openclaw/openclaw/issues/54864 | feat: add --new-session flag to openclaw agent CLI | ## Context
The `openclaw agent` CLI derives a session key from the agent name when no `--session-id` or `--session-key` is provided. This means repeated calls to the same agent reuse conversation history by default.
Tools that want isolated sessions (like workflow engines or batch scripts) currently have to generate ... | open | null | false | 1 | [] | [] | 2026-03-26T03:52:50Z | 2026-05-01T21:07:25Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | unboxed-ai | 253,808,488 | U_kgDODyDPaA | User | false |
openclaw/openclaw | 4,140,384,130 | I_kwDOQb6kR872yT-C | 54,877 | https://github.com/openclaw/openclaw/issues/54877 | https://api.github.com/repos/openclaw/openclaw/issues/54877 | [Bug]: doctor suggests incomplete sessions cleanup preview for missing transcripts | ## Summary
`openclaw doctor` reports missing recent session transcripts, but the suggested preview command does not surface the same missing entries unless `--fix-missing` is manually added.
This creates a misleading workflow for users: `doctor` says recent sessions are missing transcripts, but `openclaw sessions cle... | open | null | false | 1 | [] | [] | 2026-03-26T04:10:36Z | 2026-05-04T21:59:37Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Suidge | 34,770,064 | MDQ6VXNlcjM0NzcwMDY0 | User | false |
openclaw/openclaw | 4,140,388,324 | I_kwDOQb6kR872yU_k | 54,878 | https://github.com/openclaw/openclaw/issues/54878 | https://api.github.com/repos/openclaw/openclaw/issues/54878 | ACP child sessions spawned by a parent agent are not readable via sessions_history without cross-agent allowlist | ## Summary
Parent agents can successfully spawn ACP child sessions, but cannot read those child sessions back with `sessions_history` unless the ACP harness agent id (for example `codex`) is manually added to `tools.agentToAgent.allow`.
This makes ACP orchestration half-functional: spawn works, but result retrieval is... | open | null | false | 2 | [] | [] | 2026-03-26T04:11:26Z | 2026-05-04T21:34:50Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | 1sexywoo8 | 262,731,466 | U_kgDOD6j2yg | User | false |
openclaw/openclaw | 4,140,389,771 | I_kwDOQb6kR872yVWL | 54,879 | https://github.com/openclaw/openclaw/issues/54879 | https://api.github.com/repos/openclaw/openclaw/issues/54879 | Browser tool needs a CDP/IME text insertion action for contenteditable editors like X composer | ## Summary
The current browser tool input paths are not sufficient for some contenteditable rich editors (notably X/Twitter composer with Korean text input).
A CDP-backed text insertion action (for example using Chrome DevTools Protocol `Input.insertText`) would solve a real capability gap.
## Problem observed
On X c... | open | null | false | 1 | [] | [] | 2026-03-26T04:11:41Z | 2026-05-04T21:42:28Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | 1sexywoo8 | 262,731,466 | U_kgDOD6j2yg | User | false |
openclaw/openclaw | 4,140,409,888 | I_kwDOQb6kR872yaQg | 54,882 | https://github.com/openclaw/openclaw/issues/54882 | https://api.github.com/repos/openclaw/openclaw/issues/54882 | [Bug]: Tool display missing file path when model uses "file" parameter alias | ## Description
When the model calls the `Read`, `Write`, or `Edit` tool using the `file` parameter alias (instead of `path`, `file_path`, or `filePath`), the verbose/block-streaming tool summary shows only the bare emoji + label (e.g. `📖 Read`) without the file path detail.
The tool **executes correctly** — `normali... | open | null | false | 3 | [] | [] | 2026-03-26T04:15:16Z | 2026-05-04T13:52:37Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | mingming099 | 80,340,617 | MDQ6VXNlcjgwMzQwNjE3 | User | false |
openclaw/openclaw | 4,140,534,650 | I_kwDOQb6kR872y4t6 | 54,895 | https://github.com/openclaw/openclaw/issues/54895 | https://api.github.com/repos/openclaw/openclaw/issues/54895 | Feature: Inter-session progress streaming for sessions_send (ACP/long-running session support) | ## Summary
When `sessions_send` targets a long-running session (e.g. an ACP/Codex binding session), the caller receives a `timeout` status even though the message was successfully delivered and the target session is actively working. The caller has **no visibility** into the target session's progress (tool calls, assi... | open | null | false | 3 | [] | [] | 2026-03-26T04:39:56Z | 2026-05-04T21:42:30Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Proita | 210,555,382 | U_kgDODIzR9g | User | false |
openclaw/openclaw | 4,140,616,891 | I_kwDOQb6kR872zMy7 | 54,909 | https://github.com/openclaw/openclaw/issues/54909 | https://api.github.com/repos/openclaw/openclaw/issues/54909 | [Bug]: Telegram inline button callback_query not routed to agent — hallucination instead of tool call | ### Bug type
Regression (worked before, now fails)
### Summary
When a Telegram inline button is pressed, the callback_query is not properly routed to trigger the agent's configured http tool. Instead of following SKILL.md/SOUL.md instructions to call an external webhook via POST, the agent hallucinates a confirmatio... | open | null | false | 4 | [
"bug",
"regression"
] | [] | 2026-03-26T05:00:12Z | 2026-05-04T21:42:31Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | timt80 | 141,552,772 | U_kgDOCG_shA | User | false |
openclaw/openclaw | 4,140,670,955 | I_kwDOQb6kR872zZ_r | 54,918 | https://github.com/openclaw/openclaw/issues/54918 | https://api.github.com/repos/openclaw/openclaw/issues/54918 | 功能建议：左侧对话列表管理功能优化 | 📌 **用户功能建议**
**功能需求：左侧对话列表管理功能优化**
**需求描述：**
**1. 对话排序/分组**
- 支持拖拽调整对话顺序
- 支持按文件夹/标签分组管理对话
- 支持自定义分类
**2. 对话重命名/移动**
- 双击或右键可重命名对话
- 支持将对话移动到其他分类/文件夹
**参考：**
类似 ChatGPT、Claude 的对话管理方式
---
*来源：用户反馈 (2026-03-26)* | open | null | false | 1 | [] | [] | 2026-03-26T05:11:39Z | 2026-05-04T21:35:05Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | jsfyq917-crypto | 256,284,736 | U_kgDOD0aYQA | User | false |
openclaw/openclaw | 4,140,677,139 | I_kwDOQb6kR872zbgT | 54,919 | https://github.com/openclaw/openclaw/issues/54919 | https://api.github.com/repos/openclaw/openclaw/issues/54919 | [Feature]: agents.defaults.thinkingDefault config field not working | ### Summary
Support setting default thinking level and reasoning visibility via config (agents.defaults.thinkingDefault and agents.defaults.reasoningDefault) so new sessions inherit these preferences without manual toggling.
### Problem to solve
Currently /think and /reasoning directives work interactively to set th... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-26T05:12:55Z | 2026-05-04T04:35:52Z | 2026-05-01T18:09:06Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | Lrs-GreenBottle | 253,340,193 | U_kgDODxmqIQ | User | false |
openclaw/openclaw | 4,140,768,584 | I_kwDOQb6kR872zx1I | 54,928 | https://github.com/openclaw/openclaw/issues/54928 | https://api.github.com/repos/openclaw/openclaw/issues/54928 | Feature: support custom Exa web_search baseUrl (or fail with explicit schema/docs guidance) | ## Summary
OpenClaw's built-in Exa web search provider still appears to reject a custom `baseUrl` in `plugins.entries.exa.config.webSearch`, even on `2026.3.24`.
This makes it hard to use Exa-compatible proxies / self-hosted gateways / reverse-proxied Exa endpoints.
## What I expected
Exa should support a custom ba... | closed | completed | true | 4 | [] | [] | 2026-03-26T05:29:34Z | 2026-05-05T04:28:49Z | 2026-05-02T05:07:07Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | mrpl327 | 58,623,660 | MDQ6VXNlcjU4NjIzNjYw | User | false |
openclaw/openclaw | 4,141,150,053 | I_kwDOQb6kR8721O9l | 54,972 | https://github.com/openclaw/openclaw/issues/54972 | https://api.github.com/repos/openclaw/openclaw/issues/54972 | Feishu: CardKit streaming cards show [Interactive Card] instead of actual content when read by other agents | ## Bug Description
When an agent replies in a Feishu group chat using streaming mode (default), the reply is sent as a **CardKit interactive card**. When another agent in the same group tries to read/quote that message, it only sees `[Interactive Card]` instead of the actual text content.
## Root Cause
In `extension... | open | null | false | 2 | [] | [] | 2026-03-26T06:45:17Z | 2026-05-04T21:35:14Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | 1va7 | 81,729,546 | MDQ6VXNlcjgxNzI5NTQ2 | User | false |
openclaw/openclaw | 4,141,344,574 | I_kwDOQb6kR8721-c- | 54,996 | https://github.com/openclaw/openclaw/issues/54996 | https://api.github.com/repos/openclaw/openclaw/issues/54996 | [Feature] Local context token tracking when model API doesn't return usage data | ## Problem
Some model providers (e.g. MiniMax) do not return `usage` data in the format OpenClaw expects. As a result:
- `totalTokensFresh` stays `False` for all sessions
- `/status` always shows `Context: 0/205k (0%)` — completely unusable
- Auto-compaction never triggers because OpenClaw does not know the current c... | open | null | false | 1 | [] | [] | 2026-03-26T07:21:38Z | 2026-05-04T21:42:36Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | w000001 | 270,243,087 | U_kgDOEBuVDw | User | false |
openclaw/openclaw | 4,141,575,126 | I_kwDOQb6kR87222vW | 55,027 | https://github.com/openclaw/openclaw/issues/55027 | https://api.github.com/repos/openclaw/openclaw/issues/55027 | Feishu channel cannot do true block streaming; block chunks are dropped unless streaming card is enabled | ### Summary
Feishu currently does not seem to support true progressive/block delivery.
In practice, there are only two behaviors:
- **`channels.feishu.streaming` enabled**
Replies go through `FeishuStreamingSession`, and the interactive card displays text with a character-by-character playback effect.
- **`chann... | open | null | false | 2 | [] | [] | 2026-03-26T08:01:44Z | 2026-05-04T21:35:33Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | ZichaoLong | 27,690,850 | MDQ6VXNlcjI3NjkwODUw | User | false |
openclaw/openclaw | 4,141,719,528 | I_kwDOQb6kR8723Z_o | 55,036 | https://github.com/openclaw/openclaw/issues/55036 | https://api.github.com/repos/openclaw/openclaw/issues/55036 | [Bug]: Discord early commentary is not committed immediately and can be lost on gateway restart | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
On Discord, early assistant commentary is often not committed immediately as a standalone visible message.
Instead, it appears to remain inside the in-progress draft / reply-dispatcher pipeline until a later stage, often surfacing only toge... | open | null | false | 2 | [
"bug"
] | [] | 2026-03-26T08:23:41Z | 2026-05-04T21:35:35Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Gyu-bot | 25,432,177 | MDQ6VXNlcjI1NDMyMTc3 | User | false |
openclaw/openclaw | 4,141,816,560 | I_kwDOQb6kR8723xrw | 55,044 | https://github.com/openclaw/openclaw/issues/55044 | https://api.github.com/repos/openclaw/openclaw/issues/55044 | [Bug]: ignored an explicit stop/no-action instruction | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
OpenClaw ignored an explicit stop/no-action instruction and later injected an internal continuation message (“Continue where you left off…”). That should never happen. User intent to stop must override internal retry behavior.
### Steps to... | open | null | false | 1 | [
"bug"
] | [] | 2026-03-26T08:38:02Z | 2026-05-04T21:59:38Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | wfellacher-droid | 253,344,125 | U_kgDODxm5fQ | User | false |
openclaw/openclaw | 4,141,830,540 | I_kwDOQb6kR87231GM | 55,046 | https://github.com/openclaw/openclaw/issues/55046 | https://api.github.com/repos/openclaw/openclaw/issues/55046 | [Bug]: Media Understanding Pipeline is not working for non-visual models | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
After correctly and explicitly configuring the tools.media.image model in openclaw.json, when a model that does not support image understanding is activated in a channel, the openclaw gateway will not automatically call the configured image... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-26T08:40:32Z | 2026-04-30T04:33:16Z | 2026-04-27T20:19:48Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | MaBeitian | 119,392,741 | U_kgDOBx3J5Q | User | false |
openclaw/openclaw | 4,141,837,644 | I_kwDOQb6kR872321M | 55,047 | https://github.com/openclaw/openclaw/issues/55047 | https://api.github.com/repos/openclaw/openclaw/issues/55047 | [Bug]: controlUI options disappeared | ### Bug type
Regression (worked before, now fails)
### Summary
In previous version I could view channels, instances, sessions, agents, skills, etc in the left hand menu and have options I could edit. Now they all return a blank screen, with just the page title
### Steps to reproduce
1. opencontrolui
2. options to ... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-26T08:41:47Z | 2026-05-05T04:26:54Z | 2026-05-02T15:37:55Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | ijedgington-blip | 246,515,439 | U_kgDODrGG7w | User | false |
openclaw/openclaw | 4,142,250,595 | I_kwDOQb6kR8725bpj | 55,071 | https://github.com/openclaw/openclaw/issues/55071 | https://api.github.com/repos/openclaw/openclaw/issues/55071 | Feature Request: Skill-Driven Plugin Context Visibility (activates.plugins) | # Feature Request: Skill-Driven Plugin Context Visibility
## Situation
As OpenClaw's plugin ecosystem grows (currently 84 extensions, 75 with independent npm packages), the framework faces a fundamental scalability challenge unique to AI Agent architectures: **every registered Tool's function schema consumes LLM cont... | open | null | false | 3 | [] | [] | 2026-03-26T09:39:48Z | 2026-05-02T19:15:18Z | null | NONE | null | 20260502T210030Z | 2026-05-02T21:00:30Z | supperdsj | 10,287,320 | MDQ6VXNlcjEwMjg3MzIw | User | false |
openclaw/openclaw | 4,142,647,292 | I_kwDOQb6kR87268f8 | 55,095 | https://github.com/openclaw/openclaw/issues/55095 | https://api.github.com/repos/openclaw/openclaw/issues/55095 | [Bug]: errorCode=INVALID_REQUEST errorMessage=missing scope: operator.write conn=28b5f2bb…fa4b id=d990912d…51f5 | ### Bug type
Regression (worked before, now fails)
### Summary
python远程使用websocket连接opneclaw3.24，无法发送消息
errorCode=INVALID_REQUEST errorMessage=missing scope: operator.write conn=28b5f2bb…fa4b id=d990912d…51f5
### Steps to reproduce
<img width="680" height="57" alt="Image" src="https://github.com/user-attachments/a... | closed | not_planned | true | 2 | [
"bug"
] | [] | 2026-03-26T10:40:24Z | 2026-04-29T04:32:14Z | 2026-04-27T03:20:49Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | WangGang6656 | 67,503,773 | MDQ6VXNlcjY3NTAzNzcz | User | false |
openclaw/openclaw | 4,142,718,289 | I_kwDOQb6kR8727N1R | 55,099 | https://github.com/openclaw/openclaw/issues/55099 | https://api.github.com/repos/openclaw/openclaw/issues/55099 | Telegram: sticky IPv4-only dispatcher still triggers on macOS with /etc/hosts IPv6 override (post-#40740) | ## Description
After PR #40740 (v2026.3.11) fixed resolver-scoped dispatchers, the sticky IPv4-only dispatcher still activates under a specific but reproducible network configuration, causing all Telegram API calls to fail silently.
## Environment
- OpenClaw: 2026.3.23-2
- Node: 24.x
- macOS (LaunchAgent gateway)
- ... | open | null | false | 3 | [] | [] | 2026-03-26T10:52:16Z | 2026-05-01T20:10:45Z | null | NONE | null | 20260501T210037Z | 2026-05-01T21:00:37Z | ntlekhugov | 194,954,152 | U_kgDOC57DqA | User | false |
openclaw/openclaw | 4,143,146,115 | I_kwDOQb6kR87282SD | 55,126 | https://github.com/openclaw/openclaw/issues/55126 | https://api.github.com/repos/openclaw/openclaw/issues/55126 | memorySearch outputDimensionality not passed to OpenAI embeddings API | ## Summary
The `memorySearch.outputDimensionality` config field is accepted by the Zod schema but is not passed to the OpenAI embeddings API when `provider: "openai"` is used. This means users cannot use OpenAI's Matryoshka dimension reduction (e.g. `text-embedding-3-small` at 512 dims instead of 1536).
## Steps to R... | open | null | false | 2 | [] | [] | 2026-03-26T11:51:58Z | 2026-05-04T21:59:44Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | vdruts | 20,264,016 | MDQ6VXNlcjIwMjY0MDE2 | User | false |
openclaw/openclaw | 4,144,203,570 | I_kwDOQb6kR873A4cy | 55,196 | https://github.com/openclaw/openclaw/issues/55196 | https://api.github.com/repos/openclaw/openclaw/issues/55196 | Feature: Add brave.http diagnostics flag for logging web_search calls | ## Summary
The Brave plugin handles `web_search` calls but currently emits no dedicated diagnostics flag, making it hard to verify or audit outbound API calls to `api.search.brave.com` without raising the global log level to `debug`.
## Proposed Change
Add a `brave.http` diagnostics flag (following the existing `tel... | closed | completed | true | 2 | [] | [] | 2026-03-26T13:56:27Z | 2026-05-05T04:28:27Z | 2026-05-02T06:49:36Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | mecampbellsoup | 2,043,821 | MDQ6VXNlcjIwNDM4MjE= | User | false |
openclaw/openclaw | 4,144,479,216 | I_kwDOQb6kR873B7vw | 55,208 | https://github.com/openclaw/openclaw/issues/55208 | https://api.github.com/repos/openclaw/openclaw/issues/55208 | Feature Request: LaunchDaemon mode for macOS (bypass Local Network Privacy) | ## Problem
On macOS 15+ (Sequoia), **Local Network Privacy** restricts LaunchAgents from accessing local network IPs. The gateway gets `EHOSTUNREACH` when connecting to LAN devices (Ollama, LMStudio, etc.) even though Terminal works fine.
Ref: https://github.com/23blocks-OS/ai-maestro/issues/24
## Proposed Solution
... | closed | not_planned | true | 1 | [] | [] | 2026-03-26T14:27:42Z | 2026-05-02T04:26:59Z | 2026-04-29T22:29:48Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | TEZMan | 12,647,134 | MDQ6VXNlcjEyNjQ3MTM0 | User | false |
openclaw/openclaw | 4,221,812,661 | I_kwDOQb6kR877o7-1 | 62,872 | https://github.com/openclaw/openclaw/issues/62872 | https://api.github.com/repos/openclaw/openclaw/issues/62872 | sessions_send A2A flow ping-pong causes duplicate messages in persistent sessions | # sessions_send A2A flow ping-pong causes duplicate messages in persistent sessions
## Problem Description
When using `sessions_send` for inter-agent communication in persistent sessions, the Agent-to-Agent (A2A) flow includes a ping-pong mechanism that causes duplicate messages to appear in both sessions.
## Root C... | open | null | false | 2 | [] | [] | 2026-04-08T03:10:52Z | 2026-05-03T16:21:18Z | null | NONE | null | 20260503T180029Z | 2026-05-03T18:00:29Z | tonyhth | 90,374,544 | MDQ6VXNlcjkwMzc0NTQ0 | User | false |
openclaw/openclaw | 4,221,822,513 | I_kwDOQb6kR877o-Yx | 62,874 | https://github.com/openclaw/openclaw/issues/62874 | https://api.github.com/repos/openclaw/openclaw/issues/62874 | [Bug]: Tool calls not properly returned in tool_calls field when using local vLLM + Qwen models with tool_choice: "auto" | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When using local vLLM with Qwen models, tool calls are not placed in the tool_calls array when tool_choice defaults to "auto", causing OpenClaw Skills to fail to execute.
### Steps to reproduce
Start vLLM wit... | closed | not_planned | true | 4 | [
"bug",
"bug:behavior"
] | [] | 2026-04-08T03:13:35Z | 2026-04-30T04:35:22Z | 2026-04-27T09:32:48Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | zsjlovelike | 64,449,443 | MDQ6VXNlcjY0NDQ5NDQz | User | false |
openclaw/openclaw | 4,221,861,316 | I_kwDOQb6kR877pH3E | 62,877 | https://github.com/openclaw/openclaw/issues/62877 | https://api.github.com/repos/openclaw/openclaw/issues/62877 | BTW usage string `<side question>` placeholder eaten by markdown channels | ## Problem
`commands-btw.ts:7` defines:
```ts
const BTW_USAGE = "Usage: /btw <side question>";
```
When a user sends `/btw` with no question, this string is returned to the channel as-is. Channels that render replies as markdown (e.g. dingtalk, slack, anything HTML-based) interpret `<side question>` as an unknown HT... | open | null | false | 1 | [] | [] | 2026-04-08T03:23:52Z | 2026-05-03T09:48:48Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | BrilliantWang | 120,538,413 | U_kgDOBy9FLQ | User | false |
openclaw/openclaw | 4,221,901,656 | I_kwDOQb6kR877pRtY | 62,891 | https://github.com/openclaw/openclaw/issues/62891 | https://api.github.com/repos/openclaw/openclaw/issues/62891 | [v4.5] Session-store reconciliation on startup blocks main process, causing full fallback chain timeout for cron/LLM requests | null | open | null | false | 3 | [] | [] | 2026-04-08T03:36:14Z | 2026-05-03T16:21:20Z | null | NONE | null | 20260503T180029Z | 2026-05-03T18:00:29Z | Stache73 | 7,196,384 | MDQ6VXNlcjcxOTYzODQ= | User | false |
openclaw/openclaw | 4,221,990,179 | I_kwDOQb6kR877pnUj | 62,906 | https://github.com/openclaw/openclaw/issues/62906 | https://api.github.com/repos/openclaw/openclaw/issues/62906 | [Bug] Agent responds multiple times with identical content, causing frustrating user experience | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
SEE FULL REPORT HERE--->
[openclaw-bug-report.md](https://github.com/user-attachments/files/26558089/openclaw-bug-report.md)
## Description
### Expected Behavior
When a user interacts with OpenClaw, the agent should provi... | open | null | false | 2 | [
"bug",
"regression"
] | [] | 2026-04-08T03:59:28Z | 2026-05-03T16:21:21Z | null | NONE | null | 20260503T180029Z | 2026-05-03T18:00:29Z | taozhiyuai | 146,583,103 | U_kgDOCLyuPw | User | false |
openclaw/openclaw | 4,222,043,272 | I_kwDOQb6kR877p0SI | 62,917 | https://github.com/openclaw/openclaw/issues/62917 | https://api.github.com/repos/openclaw/openclaw/issues/62917 | fix: exec-approvals does not respect OPENCLAW_STATE_DIR, causing duplicate state-dir doctor warning | ## Summary
`exec-approvals.ts` does not respect `OPENCLAW_STATE_DIR` when resolving the path for `exec-approvals.json` and `exec-approvals.sock`. On setups that use `OPENCLAW_STATE_DIR` to relocate app data (without setting `OPENCLAW_HOME`), these two files continue to land in `~/.openclaw/` instead of the configured ... | closed | not_planned | true | 2 | [] | [] | 2026-04-08T04:11:10Z | 2026-05-02T04:27:19Z | 2026-04-29T20:05:05Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | LouisGameDev | 1,916,460 | MDQ6VXNlcjE5MTY0NjA= | User | false |
openclaw/openclaw | 4,222,222,279 | I_kwDOQb6kR877qf_H | 62,939 | https://github.com/openclaw/openclaw/issues/62939 | https://api.github.com/repos/openclaw/openclaw/issues/62939 | [Feature]: Prompt injection defense at tool result and message boundaries | ## Summary
Add structural delimiters that mark externally-sourced content (tool results, incoming messages, web fetches) as data rather than instructions, to defend against prompt injection attacks.
## Problem to solve
OpenClaw processes untrusted content from multiple surfaces: incoming user messages, file reads, w... | open | null | false | 3 | [] | [] | 2026-04-08T04:58:27Z | 2026-05-03T09:48:49Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | sarkarsaurabh27 | 1,854,562 | MDQ6VXNlcjE4NTQ1NjI= | User | false |
openclaw/openclaw | 4,222,303,575 | I_kwDOQb6kR877qz1X | 62,944 | https://github.com/openclaw/openclaw/issues/62944 | https://api.github.com/repos/openclaw/openclaw/issues/62944 | [Bug]: image tool uses hardcoded 30s timeout, ignores tools.media.models.timeoutSeconds config | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
## Environment
- OS: Ubuntu 24.04
- OpenClaw version: 20260407
- Image model: lmstudio/qwen3.5-122b-a10b
## Describe the bug
When using the `image` tool with a slow image model, the request is always
aborted after 30 second... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-04-08T05:21:10Z | 2026-04-30T04:34:50Z | 2026-04-27T10:37:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Orangewang124 | 66,901,097 | MDQ6VXNlcjY2OTAxMDk3 | User | false |
openclaw/openclaw | 4,222,359,457 | I_kwDOQb6kR877rBeh | 62,954 | https://github.com/openclaw/openclaw/issues/62954 | https://api.github.com/repos/openclaw/openclaw/issues/62954 | Heartbeat turn overwrites session's persisted model/modelProvider, leaving stale model in UI display | ## Summary
When the built-in heartbeat (`agents.defaults.heartbeat.model`) runs a turn, it correctly resolves its model override per-turn — but then the post-turn usage persistence step overwrites the session entry's `model` / `modelProvider` fields with the heartbeat model. Those fields back the UI's "current model" ... | open | null | false | 2 | [] | [] | 2026-04-08T05:35:47Z | 2026-05-03T13:21:13Z | null | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | tumbleweedlabs | 4,380,794 | MDQ6VXNlcjQzODA3OTQ= | User | false |
openclaw/openclaw | 4,222,386,214 | I_kwDOQb6kR877rIAm | 62,957 | https://github.com/openclaw/openclaw/issues/62957 | https://api.github.com/repos/openclaw/openclaw/issues/62957 | [Feature]: Remove the 16k ctx hard limit for models. | ### Summary
Remove 16k ctx window min for models
### Problem to solve
Blocking models that i would like to attempt or experiment with. I can get it force loaded but still issues.
### Proposed solution
Just remove it...noone gets hurt, switch to a warning box.
### Alternatives considered
_No response_
### Impact... | closed | completed | false | 1 | [
"enhancement"
] | [] | 2026-04-08T05:43:54Z | 2026-05-03T07:05:48Z | 2026-05-03T07:05:48Z | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | noobsmoker | 46,120,128 | MDQ6VXNlcjQ2MTIwMTI4 | User | false |
openclaw/openclaw | 4,222,414,751 | I_kwDOQb6kR877rO-f | 62,962 | https://github.com/openclaw/openclaw/issues/62962 | https://api.github.com/repos/openclaw/openclaw/issues/62962 | [Feature]: Tool security classification properties (isDestructive, isReadOnly, isConcurrencySafe) | ## Summary
Add six standard security classification properties to tool definitions so safety decisions (confirmation prompts, parallel scheduling, abort handling) are schema-driven rather than requiring per-call case analysis.
## Problem to solve
Tool definitions currently carry `ownerOnly` and participate in allow/... | open | null | false | 1 | [] | [] | 2026-04-08T05:51:54Z | 2026-05-03T06:07:14Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | sarkarsaurabh27 | 1,854,562 | MDQ6VXNlcjE4NTQ1NjI= | User | false |
openclaw/openclaw | 4,222,538,957 | I_kwDOQb6kR877rtTN | 62,976 | https://github.com/openclaw/openclaw/issues/62976 | https://api.github.com/repos/openclaw/openclaw/issues/62976 | Doctor cannot recover from invalid third-party plugin config; gateway hard-fails to start | ## Problem
A bad third-party plugin config can put OpenClaw into a hard startup failure where `openclaw doctor --fix` does not recover the system.
In practice this becomes a P0 outage: gateway won't start until users manually edit/remove plugin config entries.
## Repro (real environment)
- OpenClaw: 2026.3.24
- Add a... | closed | completed | true | 1 | [] | [] | 2026-04-08T06:24:11Z | 2026-04-30T04:32:03Z | 2026-04-28T03:51:33Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Doraemon-Claw | 263,122,318 | U_kgDOD67tjg | User | false |
openclaw/openclaw | 4,222,569,416 | I_kwDOQb6kR877r0vI | 62,981 | https://github.com/openclaw/openclaw/issues/62981 | https://api.github.com/repos/openclaw/openclaw/issues/62981 | session file locked when gateway times out and falls back to embedded runner | ## Bug
When `openclaw agent` times out at the gateway and falls back to the embedded runner, the embedded runner reuses the same `sessionId`. The gateway process still holds the session write lock, so the embedded runner fails immediately with:
```
session file locked (timeout 10000ms): pid=XXXX *.jsonl.lock
```
Thi... | closed | completed | true | 2 | [] | [] | 2026-04-08T06:31:52Z | 2026-05-02T04:27:33Z | 2026-04-29T18:38:13Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | whaili | 11,846,803 | MDQ6VXNlcjExODQ2ODAz | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.