Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 7 additions & 5 deletions docs/interactive-capabilities/capabilities.json
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@
"title": "OpenBitFun Playbook",
"origin": "https://playbook.openbitfun.com",
"source": "src/shared/interactive-capabilities/catalog.json",
"digest": "47d0b05e8a0f73d27f5f8c7a79539589d9fdaa74c8d26fcc5f5027eb4753c6b0",
"digest": "3d50d21ee69fe9a54f064e9364529f0effb58315ba85093d302cb7f6d395cfd9",
"ownerDigest": "6f0746578f3c56479f7c289bb05395dfe6917042f4d419de8c9a2f719d31f815",
"searchAcceptance": [
{
Expand Down Expand Up @@ -33371,8 +33371,8 @@
"id": "default-permission-posture",
"titleZh": "在需要确认(默认)、自动审批与完全访问三种默认权限模式间切换",
"titleEn": "Switch the default permission mode between Ask (default), Auto-approve, and Full access",
"descriptionZh": "“需要确认”对写入和命令执行保持人工把关;“自动审批”先由模型评估风险;“完全访问”需二次确认高风险提示,跳过常规询问但依然遵守产品底层硬性安全限制。",
"descriptionEn": "Ask keeps a human in the loop for writes and shell commands; Auto-approve screens calls with an AI reviewer first; Full access requires confirming a risk warning and skips routine prompts while keeping hard product safety limits intact.",
"descriptionZh": "“需要确认”默认放行普通工作区编辑、已识别的只读命令、任务调度和会话间消息,对其他命令、删除、敏感文件和发布操作保持人工把关;“自动审批”先由模型评估风险;“完全访问”需二次确认高风险提示,跳过常规询问但依然遵守产品底层硬性安全限制。",
"descriptionEn": "Ask allows ordinary workspace edits, recognized read-only commands, scheduling, and messages between sessions, while keeping other commands, deletions, sensitive files, and publishing subject to confirmation; Auto-approve screens calls with an AI reviewer first; Full access requires confirming a risk warning and skips routine prompts while keeping hard product safety limits intact.",
"control": {
"kind": "open",
"reasonCode": "unstructuredInteraction",
Expand Down Expand Up @@ -33448,12 +33448,14 @@
"tipsZh": [
"“自动审批”模式下,只有默认设为“自动审批”时下方的模型类型、超时与回退下拉框才可编辑;若在远程控制或手机端使用,建议将“不确定时”和“出错时”保留为“询问”,这样远端也能收到审批卡片而不会静默失败。",
"填写专用决策模型(Jev)的 API Key 后,记得点击输入框右侧的保存图标写入主机加密凭据库;清空密钥同理点击删除图标。",
"即使开启“完全访问”,OpenBitFun 内置的产品级硬性安全边界依然生效;项目级规则可在具体工作区中覆盖全局规则。"
"即使开启“完全访问”,OpenBitFun 内置的产品级硬性安全边界依然生效;项目级规则可在具体工作区中覆盖全局规则。",
"命令规则选择 process.execute,并用 git clone * 这样的参数模式匹配已解析的单条命令;尾部 * 不会跨过 &&、管道或重定向。其他命令和文件访问仍分别检查;无法解析的命令继续请求确认。已有 bash 规则仍可使用。"
],
"tipsEn": [
"The Auto-approve screener, timeout, and fallback controls become editable once the default mode is set to Auto-approve; keeping both fallbacks on Ask ensures mobile or Remote Connect sessions can still approve edge cases interactively.",
"After typing a dedicated Jev decision-model API key, click the Save icon beside the password box to store it in the host encrypted credential vault.",
"Even in Full access mode, hard product safety boundaries remain enforced, and project-scoped rules can still refine global defaults."
"Even in Full access mode, hard product safety boundaries remain enforced, and project-scoped rules can still refine global defaults.",
"Use process.execute with an argument pattern such as git clone * to match a parsed command. A trailing * does not span &&, pipelines, or redirections. Other commands and file accesses are checked separately; unsupported syntax still asks for approval. Existing bash rules remain usable."
],
"agentExamplesZh": [
"打开权限规则设置,我想开启自动审批",
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -28,8 +28,8 @@ Set the default safety posture for agent tool calls (Ask, Auto-approve, or Full

- **智能体可定位入口,需交互完成 / Agent opens; interaction required** · 在需要确认(默认)、自动审批与完全访问三种默认权限模式间切换
- Switch the default permission mode between Ask (default), Auto-approve, and Full access
- 说明:“需要确认”对写入和命令执行保持人工把关;“自动审批”先由模型评估风险;“完全访问”需二次确认高风险提示,跳过常规询问但依然遵守产品底层硬性安全限制。
- Why & how: Ask keeps a human in the loop for writes and shell commands; Auto-approve screens calls with an AI reviewer first; Full access requires confirming a risk warning and skips routine prompts while keeping hard product safety limits intact.
- 说明:“需要确认”默认放行普通工作区编辑、已识别的只读命令、任务调度和会话间消息,对其他命令、删除、敏感文件和发布操作保持人工把关;“自动审批”先由模型评估风险;“完全访问”需二次确认高风险提示,跳过常规询问但依然遵守产品底层硬性安全限制。
- Why & how: Ask allows ordinary workspace edits, recognized read-only commands, scheduling, and messages between sessions, while keeping other commands, deletions, sensitive files, and publishing subject to confirmation; Auto-approve screens calls with an AI reviewer first; Full access requires confirming a risk warning and skips routine prompts while keeping hard product safety limits intact.
- **智能体可定位入口,需交互完成 / Agent opens; interaction required** · 为自动审批选择聊天审查模型或专用 Jev 决策模型并加密保存凭据
- Choose a chat screener model or dedicated Jev decision model for auto-approval with encrypted credentials
- 说明:聊天模型模式默认优先选用“快速模型”(亦可改为“主模型”或任一已启用模型);专用决策模型模式默认使用 jev-latest,其 API Key 仅加密保存在当前执行主机。
Expand Down Expand Up @@ -89,6 +89,8 @@ Set the default safety posture for agent tool calls (Ask, Auto-approve, or Full
- After typing a dedicated Jev decision-model API key, click the Save icon beside the password box to store it in the host encrypted credential vault.
- 即使开启“完全访问”,OpenBitFun 内置的产品级硬性安全边界依然生效;项目级规则可在具体工作区中覆盖全局规则。
- Even in Full access mode, hard product safety boundaries remain enforced, and project-scoped rules can still refine global defaults.
- 命令规则选择 process.execute,并用 git clone * 这样的参数模式匹配已解析的单条命令;尾部 * 不会跨过 &&、管道或重定向。其他命令和文件访问仍分别检查;无法解析的命令继续请求确认。已有 bash 规则仍可使用。
- Use process.execute with an argument pattern such as git clone * to match a parsed command. A trailing * does not span &&, pipelines, or redirections. Other commands and file accesses are checked separately; unsupported syntax still asks for approval. Existing bash rules remain usable.

关联条目 / Related: `setting.tools.execution`, `setting.ai.models`, `feature.ai-assistant`, `feature.remote-connect`

Expand Down
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
{
"schemaVersion": 1,
"generatedFrom": "src/shared/interactive-capabilities/catalog.json",
"catalogDigest": "47d0b05e8a0f73d27f5f8c7a79539589d9fdaa74c8d26fcc5f5027eb4753c6b0",
"catalogDigest": "3d50d21ee69fe9a54f064e9364529f0effb58315ba85093d302cb7f6d395cfd9",
"count": 231,
"reasonCounts": {
"externalAuth": 4,
Expand Down
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
{
"schemaVersion": 2,
"generatedFrom": "src/shared/interactive-capabilities/catalog.json",
"catalogDigest": "47d0b05e8a0f73d27f5f8c7a79539589d9fdaa74c8d26fcc5f5027eb4753c6b0",
"catalogDigest": "3d50d21ee69fe9a54f064e9364529f0effb58315ba85093d302cb7f6d395cfd9",
"commandCount": 701,
"coverage": {
"commandCount": 701,
Expand Down
4 changes: 2 additions & 2 deletions src/apps/cli/tests/support/mod.rs
Original file line number Diff line number Diff line change
Expand Up @@ -573,7 +573,7 @@ fn serve_model_response(
"index": 0,
"delta": {
"role": "assistant",
"content": "DECISION=ALLOW;RISK=LOW;REASON=READ_ONLY"
"content": "DECISION=ALLOW"
},
"finish_reason": "stop"
}]
Expand Down Expand Up @@ -739,7 +739,7 @@ fn is_permission_approval_request(request: &[u8]) -> bool {
messages.iter().any(|message| {
message["content"]
.as_str()
.is_some_and(|content| content.contains("You are a permission classifier."))
.is_some_and(|content| content.starts_with("# Permission review\n\n"))
})
})
}
Expand Down
Loading
Loading