docs: align cli option behavior

docs/en/configuration/env-vars.md

@@ -1,6 +1,6 @@
 # Environment variables
 
-Kimi Code CLI uses environment variables to control a small number of runtime behaviors — relocating the data directory, turning off telemetry, and temporarily switching models without touching the config file.
+Kimi Code CLI uses environment variables to control a small number of runtime behaviors — relocating the data directory, turning off telemetry, selecting non-interactive output, and temporarily switching models without touching the config file.
 
 ::: warning Important: API keys are not configured here
 Credential variables such as `KIMI_API_KEY`, `ANTHROPIC_API_KEY`, and `OPENAI_API_KEY` are **not** read automatically from shell environment variables. Running `export KIMI_API_KEY=xxx` in the terminal does not give any provider its key — they must be written in `config.toml` under `[providers.<name>]` or the `[providers.<name>.env]` sub-table.
@@ -32,6 +32,17 @@ Set to `1` to turn off anonymous telemetry reporting (also accepts `true`, `yes`
 export KIMI_DISABLE_TELEMETRY=1
 ```
 
+### `KIMI_MODEL_OUTPUT_FORMAT`
+
+Sets the default output format for non-interactive `-p` / `--prompt` mode. Accepted values are `text` and `stream-json`:
+
+```sh
+export KIMI_MODEL_OUTPUT_FORMAT="stream-json"
+kimi -p "List changed files"
+```
+
+An explicit `--output-format` flag has higher priority. The variable is ignored outside prompt mode, and an invalid value causes startup to fail. Despite its name, it does not require `KIMI_MODEL_NAME`.
+
 ### `KIMI_MODEL_*` family
 
 Switch models temporarily without modifying `config.toml` — when `KIMI_MODEL_NAME` is set, the CLI synthesizes a temporary provider in memory; the change does not persist after restart. See [Define a model from environment variables](#define-a-model-from-environment-variables-kimi-model).

docs/en/configuration/overrides.md

@@ -66,9 +66,11 @@ Options passed at startup have the highest priority and apply only to the curren
 Mutual exclusion rules (startup fails if violated):
 
 - `--output-format` can only be used with `-p`
-- `--prompt` cannot be combined with `--yolo` or `--plan`
+- `--prompt` cannot be combined with `--yolo`, `--auto`, or `--plan`
 - `--continue` and `--session` cannot be used together
-- In non-prompt mode, `--yolo` and `--plan` cannot be combined with `--continue` or `--session`
+- `--yolo` and `--auto` cannot be used together
+
+When resuming a session, `--auto`, `--yolo`, or `--plan` can temporarily override its saved permission or plan mode.
 
 ::: tip
 `--skills-dir` is a one-shot replacement that only affects the current startup. To persistently add search directories, write `extra_skill_dirs` in `config.toml` (see [Agent Skills](../customization/skills.md)).
@@ -89,10 +91,10 @@ KIMI_CODE_HOME="$PWD/.kimi-sandbox" kimi
 KIMI_API_KEY = "sk-test"
 ```
 
-**Skip approval for batch tasks**:
+**Run a batch task non-interactively** — prompt mode uses auto permission automatically:
 
 ```sh
-kimi --yolo -p "Batch rename the following files..."
+kimi -p "Batch rename the following files..."
 ```
 
 **Enter Plan mode temporarily** (to make it permanent, set `default_plan_mode = true` in the config file):

docs/en/reference/kimi-command.md

@@ -19,7 +19,7 @@ All flags are optional — run `kimi` directly to enter an interactive session:
 | `--continue` | `-c` | Continue the most recent session in the current working directory, without specifying an ID manually |
 | `--model <model>` | `-m` | Specify a model alias for this launch. When omitted, new sessions use `default_model` from the config file |
 | `--prompt <prompt>` | `-p` | Run a single prompt non-interactively and stream the Assistant output to stdout. This mode does not open the TUI |
-| `--output-format <format>` | | Set the non-interactive output format; supports `text` and `stream-json`. Can only be used with `--prompt`; defaults to `text` |
+| `--output-format <format>` | | Set the non-interactive output format; supports `text` and `stream-json`. Can only be used with `--prompt`. Overrides `KIMI_MODEL_OUTPUT_FORMAT`; otherwise defaults to `text` |
 | `--yolo` | `-y` | Auto-approve regular tool calls, skipping approval requests |
 | `--auto` | | Start with auto permission mode; tool approvals are handled automatically and the Agent will not ask the user questions |
 | `--plan` | | Start a new session in Plan mode — the AI will prioritize read-only tools for exploration and planning |
@@ -129,6 +129,8 @@ When you need to parse output programmatically, use the `stream-json` format —
 kimi -p "List changed files" --output-format stream-json
 ```
 
+For scripts that always use JSONL output, set [`KIMI_MODEL_OUTPUT_FORMAT`](../configuration/env-vars.md#kimi-model-output-format). An explicit `--output-format` flag still takes priority.
+
 In `stream-json` mode, regular replies produce an Assistant message; when the model calls a tool, an Assistant message with `tool_calls` is emitted first, followed by the corresponding Tool message, then subsequent Assistant messages. Thinking content is not written to JSONL; tool progress and "resuming session" notices are still written to stderr.
 
 ## Subcommands

docs/zh/configuration/env-vars.md

@@ -1,6 +1,6 @@
 # 环境变量
 
-Kimi Code CLI 通过环境变量控制少数运行时行为——迁移数据目录、关闭遥测、不改配置文件临时切换模型。
+Kimi Code CLI 通过环境变量控制少数运行时行为——迁移数据目录、关闭遥测、选择非交互输出格式，以及不改配置文件临时切换模型。
 
 ::: warning 重要：API 密钥不在这里配置
 `KIMI_API_KEY`、`ANTHROPIC_API_KEY`、`OPENAI_API_KEY` 等密钥变量**不会**从 shell 环境变量自动读取。在终端里 `export KIMI_API_KEY=xxx` 不会让任何供应商获得密钥——必须写在 `config.toml` 的 `[providers.<name>]` 段或 `[providers.<name>.env]` 子表里。
@@ -32,6 +32,17 @@ export KIMI_CODE_HOME="/path/to/custom/kimi-code"
 export KIMI_DISABLE_TELEMETRY=1
 ```
 
+### `KIMI_MODEL_OUTPUT_FORMAT`
+
+设置非交互 `-p` / `--prompt` 模式的默认输出格式。合法值为 `text` 和 `stream-json`：
+
+```sh
+export KIMI_MODEL_OUTPUT_FORMAT="stream-json"
+kimi -p "列出已修改的文件"
+```
+
+显式传入的 `--output-format` flag 优先级更高。该变量在非 prompt 模式下会被忽略，值不合法时启动会失败。虽然名称以 `KIMI_MODEL_` 开头，但它不要求设置 `KIMI_MODEL_NAME`。
+
 ### `KIMI_MODEL_*` 系列
 
 不修改 `config.toml` 临时切换模型——设置 `KIMI_MODEL_NAME` 后，CLI 在内存里合成一个临时供应商，重启后失效。详见[用环境变量定义模型](#用环境变量定义模型-kimi-model)。

docs/zh/configuration/overrides.md

@@ -66,9 +66,11 @@ Kimi Code CLI 有三个地方可以影响运行参数：配置文件、命令行
 互斥规则（违反时启动报错）：
 
 - `--output-format` 只能配合 `-p` 使用
-- `--prompt` 不能同时用 `--yolo` 或 `--plan`
+- `--prompt` 不能同时用 `--yolo`、`--auto` 或 `--plan`
 - `--continue` 和 `--session` 不能同时用
-- 非 prompt 模式下，`--yolo` 和 `--plan` 不能配合 `--continue` 或 `--session`
+- `--yolo` 和 `--auto` 不能同时使用
+
+恢复会话时，可以通过 `--auto`、`--yolo` 或 `--plan` 临时覆盖会话保存的权限或 Plan 模式。
 
 ::: tip
 `--skills-dir` 是一次性替换，只影响本次启动。如需长期追加搜索目录，在 `config.toml` 里写 `extra_skill_dirs`（详见 [Agent Skills](../customization/skills.md)）。
@@ -89,10 +91,10 @@ KIMI_CODE_HOME="$PWD/.kimi-sandbox" kimi
 KIMI_API_KEY = "sk-test"
 ```
 
-**跳过审批运行批处理任务**：
+**以非交互方式运行批处理任务**——prompt 模式会自动使用 auto 权限：
 
 ```sh
-kimi --yolo -p "批量重命名以下文件..."
+kimi -p "批量重命名以下文件..."
 ```
 
 **临时进入 Plan 模式**（若想永久生效，在配置文件设 `default_plan_mode = true`）：

docs/zh/reference/kimi-command.md

@@ -19,7 +19,7 @@ kimi <subcommand> [options]
 | `--continue` | `-c` | 继续当前工作目录下最近一次的会话，无需手动指定 ID |
 | `--model <model>` | `-m` | 为本次启动指定模型别名。省略时新会话使用配置文件中的 `default_model` |
 | `--prompt <prompt>` | `-p` | 非交互执行单次 prompt，并把 Assistant 输出流式写到 stdout。该模式不会打开 TUI |
-| `--output-format <format>` | | 设置非交互输出格式，支持 `text` 与 `stream-json`。仅可与 `--prompt` 一起使用，默认 `text` |
+| `--output-format <format>` | | 设置非交互输出格式，支持 `text` 与 `stream-json`。仅可与 `--prompt` 一起使用。显式 flag 优先于 `KIMI_MODEL_OUTPUT_FORMAT`，后者未设置时默认 `text` |
 | `--yolo` | `-y` | 自动批准普通工具调用，跳过审批请求 |
 | `--auto` | | 以 auto 权限模式启动；工具审批自动处理，Agent 不会向用户提问 |
 | `--plan` | | 以 Plan 模式启动新会话，AI 会优先使用只读工具进行探索和规划 |
@@ -129,6 +129,8 @@ kimi -m kimi-code/kimi-for-coding -p "Explain the latest diff"
 kimi -p "List changed files" --output-format stream-json
 ```
 
+需要让脚本始终使用 JSONL 输出时，可以设置 [`KIMI_MODEL_OUTPUT_FORMAT`](../configuration/env-vars.md#kimi-model-output-format)。显式传入的 `--output-format` flag 仍然具有更高优先级。
+
 `stream-json` 模式下，普通回复输出 Assistant 消息；模型调用工具时，先输出带 `tool_calls` 的 Assistant 消息，再输出对应的 Tool 消息，最后继续输出后续 Assistant 消息。thinking 内容不会写入 JSONL；工具进度和恢复会话提示仍写到 stderr。
 
 ## 子命令