Sharkly × IM development workflow
The team discusses a topic in Feishu, WeCom, or DingTalk. Once the discussion is clear, mention the Sharkly Agent to organize the discussion, create a requirement task, and hand it to a development Agent for coding, testing, and human acceptance.
A real workflow
A requirement starts with a topic in the team group. Product, design, and engineering keep adding context to the discussion. When work should begin, an explicit mention turns the complete discussion into accountable, trackable delivery.
Product posts the topic, then design and engineering add constraints, options, and open questions around it. Whether the team uses Feishu, WeCom, or DingTalk, the discussion stays focused on one subject instead of scattering through the group feed.
The user explicitly asks the Agent to summarize the current topic, capture goals, scope, open questions, and acceptance criteria, create a requirement task, and assign the development Agent. Group messages do not become tasks unless the team asks for it.
The Agent reads the task context shaped from the discussion, repository rules, and acceptance criteria, then works in an isolated workspace. Progress, changed files, blockers, and logs return to the task instead of disappearing into a private terminal.
The Agent runs the required typecheck, tests, and build, then attaches the change summary, verification results, and known limits to the task. The task moves to Human Review, where a person accepts it or requests another pass.
Workflow outcomes
Sharkly connects the discussion context, task, execution record, and human decision. The team sees the complete path from discussion to delivery, not only the final answer copied out of a private AI window.
Decisions around one topic become a task with goals, boundaries, and acceptance criteria.
Ownership, progress, and blockers return to the same task the team can follow.
Code, test results, and known limits arrive together so a person can make the final call.
This workflow does not promise an automatic merge or deployment. Sharkly brings inspectable results to the user and stops automation at the right human boundary.
Getting started
Prepare two Agents with clear roles, then place the requirement entry point inside the Feishu, WeCom, or DingTalk group your team already uses.
Create a requirement Agent in Sharkly, then connect it to the Feishu, WeCom, or DingTalk group your team uses.
Connect its repository and Runtime, then define the coding, testing, and reporting rules it must follow.
Mention the requirement Agent and ask it to summarize the current discussion, create the task, and assign the development Agent.
FAQ
No. Messages enter the connected Agent conversation. A task is created only when a user explicitly asks the Agent under the relevant topic to summarize the discussion and create one.
Teams can use Feishu, WeCom, or DingTalk based on their existing communication habits. The workflow does not depend on one IM product; the important part is carrying the current topic into the task.
No. The recommended setup uses one Agent for communication and task shaping and another for code execution. Teams can adjust the roles to match their workflow.
This workflow stops at human acceptance. The Agent returns the code changes, summary, and test evidence, then a user decides whether to accept, request changes, or continue with a merge.