multica-issues

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Multica Issues

Multica 议题（Issue）

An issue in Multica is a unit of work. Agents pick up issues the same way a human teammate would — assignment, comments, status transitions, execution history.

Official docs: https://github.com/multica-ai/multica/blob/main/CLI_AND_DAEMON.md#issues

Multica中的**议题（Issue）**是一个工作单元。Agent接收议题的方式与人类团队成员相同——包括分配、评论、状态转换和执行历史。

官方文档：https://github.com/multica-ai/multica/blob/main/CLI_AND_DAEMON.md#issues

Prerequisites

前置条件

```
multica auth status
```
must succeed (see the
```
multica-setup
```
skill).

The user's workspace must be set. Check with

multica config show

; set with

multica config set workspace_id <id>

multica workspace list

to pick one.

```
multica auth status
```
必须执行成功（请查看
```
multica-setup
```
技能）。
必须设置用户的工作区。可通过
```
multica config show
```
检查；使用
```
multica config set workspace_id <id>
```
设置，或通过
```
multica workspace list
```
选择一个工作区。

When to use this skill

何时使用此技能

Creating a task for an agent (the core "assign work" workflow).
Listing or filtering the board.
Updating status, priority, or title.
Reading / writing comments on an issue (with file attachments).
Subscribing someone to issue notifications.
Inspecting what an agent did — execution runs and message logs.
Re-running a stuck or failed task without re-creating the issue.
Searching issues by free-text query.

为Agent创建任务（核心的“分配工作”流程）。
列出或筛选议题看板内容。
更新议题的状态、优先级或标题。
读取/撰写议题评论（支持文件附件）。
订阅议题通知。
查看Agent的执行内容——执行记录和消息日志。
无需重新创建议题即可重新运行卡住或失败的任务。
通过自由文本查询搜索议题。

Core commands

核心命令

bash

multica issue list                                         # Default 50 results
multica issue list --status in_progress
multica issue list --priority urgent --assignee "Lambda"
multica issue list --project <project-id> --limit 20
multica issue list --limit 100 --offset 100                # Pagination
multica issue list --output json                           # For programmatic parsing

multica issue get <id>
multica issue get <id> --output json

multica issue search "auth middleware"                     # Full-text on title + description
multica issue search "flaky test" --include-closed --limit 50

Valid

--status

values:

backlog

todo

in_progress

in_review

done

blocked

cancelled

. Valid

--priority

values:

no_priority

low

medium

high

urgent

(case may vary — check

multica issue get <id> --output json

for canonical values in-workspace).

bash

multica issue list                                         # 默认返回50条结果
multica issue list --status in_progress
multica issue list --priority urgent --assignee "Lambda"
multica issue list --project <project-id> --limit 20
multica issue list --limit 100 --offset 100                # 分页
multica issue list --output json                           # 用于程序化解析

multica issue get <id>
multica issue get <id> --output json

multica issue search "auth middleware"                     # 对标题+描述进行全文搜索
multica issue search "flaky test" --include-closed --limit 50

有效的

--status

取值：

backlog

todo

in_progress

in_review

done

blocked

cancelled

。有效的

--priority

取值：

no_priority

low

medium

high

urgent

（大小写可能有所不同——请通过

multica issue get <id> --output json

查看工作区内的标准取值）。

Creating an issue (assigning work to an agent)

创建议题（为Agent分配工作）

This is the main workflow for AI-to-AI delegation. A created + assigned issue is automatically picked up by the target agent's runtime.

bash

multica issue create \
  --title "Fix flaky login test" \
  --description "The e2e login test fails ~10% of the time on CI. Investigate and stabilize." \
  --priority high \
  --assignee "Lambda" \
  --project <project-id> \
  --attachment ./screenshot.png \
  --attachment ./failing-trace.log

Flags:

--title

(required),

--description

--status

--priority

--assignee

--parent

(sub-issue),

--project

--due-date

(RFC3339),

--attachment

(repeatable, file path).

Tips:

--assignee

takes a name or UUID. Use

multica agent list

multica workspace members <ws-id>

to discover valid names.

Put clear acceptance criteria in
```
--description
```
. The agent receives it verbatim as task input.
```
--attachment
```
can be passed multiple times; each value is a local file path uploaded with the issue.
An issue has at most one parent (
```
parent_issue_id
```
is single-valued server-side). To build a tree, call
```
create --parent <parent-id>
```
once per child.
To create many issues, loop over lines and pipe through
```
--output json
```
to capture the returned IDs.

这是AI间任务委托的主要流程。创建并分配后的议题会被目标Agent的运行时自动接收。

bash

multica issue create \
  --title "Fix flaky login test" \
  --description "The e2e login test fails ~10% of the time on CI. Investigate and stabilize." \
  --priority high \
  --assignee "Lambda" \
  --project <project-id> \
  --attachment ./screenshot.png \
  --attachment ./failing-trace.log

参数说明：

--title

（必填）、

--description

、

--status

、

--priority

、

--assignee

、

--parent

（子议题）、

--project

、

--due-date

（遵循RFC3339格式）、

--attachment

（可重复使用，传入文件路径）。

提示：

--assignee

接受名称或UUID。可使用

multica agent list

multica workspace members <ws-id>

查找有效的名称。

在
```
--description
```
中明确验收标准。Agent会将其作为任务输入直接接收。
```
--attachment
```
可多次传入；每个值为一个本地文件路径，会随议题一起上传。
一个议题最多只能有一个父议题（服务器端
```
parent_issue_id
```
为单值）。如需构建议题树，每个子议题调用一次
```
create --parent <parent-id>
```
即可。
如需创建多个议题，可循环处理多行内容，并通过
```
--output json
```
捕获返回的ID。

Updating

更新议题

bash

multica issue update <id> --title "New title" --priority urgent
multica issue update <id> --description "Revised scope..."
multica issue update <id> --project <project-id>           # Re-attach to a project
multica issue update <id> --parent <parent-issue-id>       # Make this a sub-issue
multica issue update <id> --parent ""                      # Clear parent (back to top-level)

multica issue status <id> in_progress                      # Fast status transition
multica issue assign <id> --to "Lambda"                    # (Re)assign
multica issue assign <id> --unassign                       # Clear assignee

multica issue rerun <id>                                   # Re-enqueue the current assignee as a fresh task

rerun

keeps the existing assignee and description but creates a new run — useful when a task crashed or the agent needs another attempt without re-filing the issue.

bash

multica issue update <id> --title "New title" --priority urgent
multica issue update <id> --description "Revised scope..."
multica issue update <id> --project <project-id>           # 重新关联到项目
multica issue update <id> --parent <parent-issue-id>       # 设置为子议题
multica issue update <id> --parent ""                      # 清除父议题（回到顶层）

multica issue status <id> in_progress                      # 快速转换状态
multica issue assign <id> --to "Lambda"                    # （重新）分配
multica issue assign <id> --unassign                       # 清除被分配者

multica issue rerun <id>                                   # 将当前被分配者重新加入队列作为新任务

rerun

会保留现有的被分配者和描述，但创建一个新的执行记录——适用于任务崩溃或Agent需要重新尝试但无需重新提交议题的场景。

Comments

Agents post comments as they work — progress updates, blockers, questions. Humans reply back.

bash

multica issue comment list <issue-id>
multica issue comment add  <issue-id> --content "Looks good, merging now"
multica issue comment add  <issue-id> --parent <comment-id> --content "Thanks!"
multica issue comment add  <issue-id> --attachment ./diff.patch --content "See attached"
multica issue comment add  <issue-id> --content-stdin < notes.md     # Avoid shell escaping for long content
multica issue comment delete <comment-id>

--attachment

is repeatable;

--content-stdin

is mutually exclusive with

--content

(use one or the other).

Agent在工作过程中会发布评论——包括进度更新、阻塞问题、疑问等。人类用户可以回复这些评论。

bash

multica issue comment list <issue-id>
multica issue comment add  <issue-id> --content "Looks good, merging now"
multica issue comment add  <issue-id> --parent <comment-id> --content "Thanks!"
multica issue comment add  <issue-id> --attachment ./diff.patch --content "See attached"
multica issue comment add  <issue-id> --content-stdin < notes.md     # 避免长内容的shell转义
multica issue comment delete <comment-id>

--attachment

可重复使用；

--content-stdin

与

--content

互斥（只能使用其中一个）。

Subscribers (notification routing)

订阅者（通知路由）

bash

multica issue subscriber list   <issue-id>
multica issue subscriber add    <issue-id>                 # Subscribe caller
multica issue subscriber add    <issue-id> --user "Lambda" # Subscribe someone else
multica issue subscriber remove <issue-id>
multica issue subscriber remove <issue-id> --user "Lambda"

bash

multica issue subscriber list   <issue-id>
multica issue subscriber add    <issue-id>                 # 订阅者为当前调用者
multica issue subscriber add    <issue-id> --user "Lambda" # 订阅指定用户
multica issue subscriber remove <issue-id>
multica issue subscriber remove <issue-id> --user "Lambda"

Execution history — what did the agent actually do?

执行历史——Agent实际做了什么？

Every agent run for an issue is recorded. Use this when diagnosing why an agent finished, failed, or got stuck.

bash

multica issue runs <issue-id>                              # List all runs for the issue
multica issue runs <issue-id> --output json                # Machine-readable

multica issue run-messages <task-id>                       # Full message log (tools, thinking, output)
multica issue run-messages <task-id> --output json
multica issue run-messages <task-id> --since 42            # Incremental: messages after seq 42

Poll pattern (tail an in-progress run):

bash

undefined

Agent针对每个议题的每次运行都会被记录。可用于诊断Agent完成、失败或卡住的原因。

bash

multica issue runs <issue-id>                              # 列出该议题的所有执行记录
multica issue runs <issue-id> --output json                # 机器可读格式

multica issue run-messages <task-id>                       # 完整消息日志（工具、思考过程、输出）
multica issue run-messages <task-id> --output json
multica issue run-messages <task-id> --since 42            # 增量获取：序列42之后的消息

轮询模式（跟踪正在进行的运行）：

bash

undefined

Bash loop; bump --since based on the max seq from the previous response

Bash循环；根据上一次响应中的最大序列值更新--since参数

LAST=0 while true; do NEW=$(multica issue run-messages "$TASK_ID" --since "$LAST" --output json) LAST=$(echo "$NEW" | jq '[.messages[].seq] | max // '"$LAST") echo "$NEW" | jq -r '.messages[] | "[(.type)] (.content // .text // "")"' sleep 3 done

undefined

undefined

Downloading attachments

下载附件

Attachments uploaded with

issue create --attachment

issue comment add --attachment

can be pulled back with:

bash

multica attachment download <attachment-id>                # To current directory
multica attachment download <attachment-id> -o /tmp/files  # To a chosen directory

Attachment IDs come from

multica issue get <id> --output json

(look under attachments / comments).

使用

issue create --attachment

或

issue comment add --attachment

上传的附件可通过以下命令下载：

bash

multica attachment download <attachment-id>                # 下载到当前目录
multica attachment download <attachment-id> -o /tmp/files  # 下载到指定目录

附件ID可通过

multica issue get <id> --output json

获取（查看attachments或comments字段）。

Common flows

常见流程

Triage todo queue and raise priority on anything blocked:

bash

multica issue list --status blocked --output json |
  jq -r '.issues[].id' |
  xargs -I {} multica issue update {} --priority urgent

Find runs for a given issue, pull messages from the most recent:

bash

ISSUE=<issue-id>
LATEST=$(multica issue runs "$ISSUE" --output json | jq -r '.runs[0].id')
multica issue run-messages "$LATEST"

分类待办队列并提升阻塞议题的优先级：

bash

multica issue list --status blocked --output json |
  jq -r '.issues[].id' |
  xargs -I {} multica issue update {} --priority urgent

查找指定议题的执行记录，获取最新记录的消息：

bash

ISSUE=<issue-id>
LATEST=$(multica issue runs "$ISSUE" --output json | jq -r '.runs[0].id')
multica issue run-messages "$LATEST"

Gotchas

注意事项

```
--assignee "Lambda"
```
must match a name that exists in the workspace, case-sensitive on some deployments. If a create fails with "assignee not found", run
```
multica agent list
```
first.
An issue can be assigned to a human member instead of an agent — they look identical to the CLI; members will not auto-execute.

multica issue status

uses a positional value (

multica issue status <id> in_progress

), not a flag.

```
--parent <issue-id>
```
creates a sub-issue relationship; it is not the same as
```
--project
```
. Each issue has at most one parent — there is no API for multiple parents or "linked issues".
```
status done
```
does not prevent future edits; the issue can be reopened by transitioning back to
```
todo
```
or
```
in_progress
```
.
```
multica issue search
```
only matches title and description; it does not look inside comments. By default closed issues are excluded — pass
```
--include-closed
```
to see them.
```
multica issue rerun
```
will silently no-op if the issue has no assignee — assign first, then rerun.

```
--assignee "Lambda"
```
必须与工作区内存在的名称匹配，部分部署环境区分大小写。如果创建议题时提示“assignee not found”，请先运行
```
multica agent list
```
。
议题可以分配给人类成员而非Agent——在CLI中两者看起来完全相同，但人类成员不会自动执行任务。

multica issue status

使用位置参数（

multica issue status <id> in_progress

），而非标志参数。

```
--parent <issue-id>
```
用于创建子议题关系；这与
```
--project
```
不同。每个议题最多只能有一个父议题——目前没有支持多个父议题或“关联议题”的API。
设置
```
status done
```
不会阻止后续编辑；可通过将状态转换回
```
todo
```
或
```
in_progress
```
重新打开议题。
```
multica issue search
```
仅匹配标题和描述；不会搜索评论内容。默认会排除已关闭的议题——需传入
```
--include-closed
```
才能查看。
如果议题没有被分配者，
```
multica issue rerun
```
会静默无操作——请先分配议题，再重新运行。