zentao

ZenTao AI Assistant

When to use this skill

Enable this skill when the user wants you to query todos in ZenTao on their behalf, summarize team items, create tasks, transition statuses, or produce morning meeting/weekly report material.

Prefer to use it as a "process-oriented skill" rather than a "command echo manual."

Output contract

• By default, directly call the underlying capabilities to complete the task; do not return zentao-cli ... commands to the user as the final reply.
• Only when the user explicitly asks "give me the command," "give me a CLI example," or "I want to execute/troubleshoot manually" should you read and cite cli-examples.md。
• Before any capability call, first read auth-and-precheck.md。
• When write operations are involved, first read write-workflows.md。
• When queries, statistics, morning meetings, weekly reports, or risk scans are involved, first read query-routing.md。
• When a compact version of the "create / my / manage" decision hints is needed, then read decision-rules.md。
• For query-type tasks, prioritize returning a natural language summary, then provide a structured list.
• For management-type results, prioritize outputting Chinese-friendly fields; avoid throwing internal field names directly at the user.
• If it is only to confirm that this skill is available, directly complete the query or operation and report the result; do not echo commands.

Layered workflow

Always think in the following 4-layer order:

1. 鉴权与预检层 First check token / sid / local login state / team cache / necessary project and requirement context. By default, prioritize using the logged-in token + sid; if authentication is invalid, then use the locally saved login information to silently log in again once, and refresh the local configuration.

2. 意图路由层 First determine whether this is:

   • write operation
   • personal query
   • management summary
   • specialized statistics

3. 能力执行层 Then choose the most appropriate one from the nine core capabilities to execute; do not force all problems into manage。

4. 结果核查层 After a write operation is completed, read back the result; after a query is completed, explain the perspective, statistical scope, and filter conditions.

Nine-core-tools map

A. Queries and Insights

1. Global todo perspective

• Applicable: viewing my own, a certain person's, or a certain team's tasks/requirements/Bugs.
• Main capabilities:my、manage
• Key distinction:
  • my = official assignment / personal territory
  • manage = management summary / team overview

2. Intelligent link parsing

• Applicable: the user directly sends a task, requirement, or Bug link.
• Main capabilities:view
• Key rules:
  • When there is a link, prioritize parsing the link first; do not first ask the user to supplement the ID
  • If the user asks "is it closed-loop," "is it already completed," or "what exactly is the current status of this task," prioritize using view
  • If it is identified as a multi-person parallel task, you must output both the "individual completion state" and the "overall closed-loop state"
  • Do not draw a conclusion based only on the overall doing/wait/done in one sentence

3. Automatic morning meeting broadcast

• Applicable: today's morning meeting, risk broadcast, due today/tomorrow, overdue list.
• Main capabilities:morning-check
• Key rules:
  • This is specialized statistics; do not degrade it into an ordinary manage
  • The morning report outputs an "item pool," which may simultaneously include tasks / requirements / Bugs
  • Do not describe all overdue items in the morning report as "overdue tasks"
  • If the user wants to verify "how many overdue tasks a certain person currently has exactly", switch to the task scope for a separate review

4. Pre-dispatch workload reference radar

• Applicable: Before dispatching, determine who is relatively free and who already has too many tasks.
• Main capabilities:load
• Key rule: This is reference information and does not automatically reject dispatch on the user's behalf.

5. Stalled ticket investigation

• Applicable: Investigate in-progress items that have not been updated for a long time.
• Main capabilities:stagnant
• Key rule: The core scope is "stall days + last update time", not simple status filtering.

6. Automated weekly report summary

• Applicable: Team weekly reports, this week's delivery summary, this week's pending items.
• Main capabilities:weekly-synthesis
• Key rule: This is weekly report statistics, not equivalent to an ordinary management aggregate view.

B. Writing and execution

7. Conversational rapid task creation

• Applicable: Ordinary task creation, splitting tasks from requirements, multi-person task creation.
• Main capabilities:task create, execute query/create
• Key rules:
  • First determine whether it is associated with a requirement
  • First verify the project and execution ownership
  • Multiple members default to multi-person parallel

8. Conversational quick work logging

• Applicable: Register time spent, supplement work notes.
• Main capabilities:task effort
• Key rule: The currently stable capability is "time spent + optional notes".

9. Minimal status transition

• Applicable: Start, complete, close, transfer, find a task by name and then transition it.
• Main capabilities:task update、story update、bug update、task find
• Key rule: First find the correct object, then perform the status transition or transfer.

C. Basic auxiliary capabilities

10. Team roster cache

• Applicable: Fixed teams repeatedly doing morning meetings, weekly reports, workload checks, and management aggregation.
• Main capabilities:team save、team list、team show、team delete
• Key rules:
  • For fixed teams, prioritize saving the team alias first
  • Subsequently prioritize reuse --team-name

11. Automatic Chinese mapping

• Applicable: The user directly inputs a real Chinese name instead of a system account.
• Main capabilities: Member mapping, assignment mapping, team member mapping
• Key rules:
  • Chinese names should be automatically mapped to the underlying account first
  • Both write operations and queries should perform mapping first, then continue execution

12. Human-friendly link and web page jump recognition

• Applicable: The user provides a large block of text, a task name, or a task ID and wants to quickly locate the ZenTao web page.
• Main capabilities:view、task find
• Key rules:
  • When a ZenTao link appears in the text, prioritize parsing it directly
  • When the user only provides a task name, the task can be searched first, then a clickable web link returned
  • When the user only provides a task ID, the corresponding web link can be returned directly

Required references

1. Authentication and pre-checks

First read auth-and-precheck.md

2. Write operation process

For task creation, requirement splitting, transfer, transition, and work logging, first read write-workflows.md

3. Query and statistics routing

Involving my、manage、view、morning-check、load、stagnant、weekly-synthesis, first read query-routing.md

4. Compact decision supplement

Read this only when a simplified supplement is needed. decision-rules.md

When to read cli-examples.md

Read it only in the following scenarios. cli-examples.md：

• The user explicitly requests CLI commands or manual operation steps.
• Local integration debugging, acceptance testing, or troubleshooting is needed for developers.
• A parameter name, command combination, or help example needs to be verified.

Do not return the commands in this file directly to the user by default.

Current limitations

• task effort --taskId <id> --desc "..." In this form, which only writes a description without filling in time spent, the page history will not reliably record a comment in the current Zentao environment.
• story update --storyId <id> --status active It has not yet taken effect reliably in the current Zentao environment.

Installation note

If the user explicitly requests installation or manual verification, you may prompt them to install and log in first:

• Installation package:@chenish/zentao-mcp-agent
• Complete login authorization before first use.
• For the specific command syntax, see cli-examples.md