How a Lightweight Bot‑Powered Tool Eliminates Manual To‑Do Lists in Test Projects

Part 1 – Dynamic Project‑Management Rules

Test teams often become “human to‑do lists”. The challenge is that project management must adapt to different scales (small, medium, large) and to internal vs external collaboration.

Small (≈3 people, 3 days): lightweight checks.

Medium: speed + stability, add technical review and joint‑day meetings.

Large: controllability and alignment, add daily stand‑ups and detailed integration plans.

A rule system was built around four dimensions (who, when, what, deliverable) and three solution directions:

Control granularity by project scale.

Distinguish node differences by project type (client‑side, external‑partner).

Define concrete actions for each dimension.

Example stand‑up record: overall progress, risk level, reason for issues, and a Todo list (item, owner, deadline).

Part 2 – Making the Rules Executable

Static documents fail because of memory load, execution cost, and lack of supervision. The solution is a bot‑driven tool that lives inside the Feishu (Lark) group chat.

Architecture

Interaction layer – Feishu group bot:

Smart bot parses @‑mentions, extracts intent.

Ordinary bot sends scheduled notifications.

Conversion layer – forwards messages to Coze and encapsulates code logic for group pushes.

Execution layer (Coze automation) :

Large‑model intent detection, key‑data extraction, validation, and update of a project‑management spreadsheet.

Timed triggers evaluate project timelines and push pre‑reminders, progress alerts, and result verification.

Custom Coze plugins integrate with Feishu Sheets, TAPD, static‑scan and coverage services.

Core Data Tables

Project Overview Table : project ID, name, link to management sheet, TAPD address, bot IDs.

Project Management Table : personnel, timelines, key tasks, status, requirements, owners.

Key Features

Automatic onboarding – standardized group input creates a row in the overview table.

Auto‑update – user messages automatically modify the management table.

Timed reminders – tasks trigger @‑mentions to responsible persons.

Tool empowerment – APIs fetch TAPD bugs, coverage, static‑scan results; Coze processes and pushes them.

Automatic closure – when TAPD status shows “released”, the corresponding item is marked completed.

Typical Interactions

Q: Why two bots? The smart bot can only respond when @‑mentioned; the ordinary bot handles proactive scheduled pushes. Q: Why not use Feishu Docs? Docs have limited automation and weak Coze integration; the bot + Coze approach offers mature plugin development.

Reference APIs

Feishu smart bot API: https://developer.work.weixin.qq.com/document/path/100719

Feishu ordinary group bot API: https://developer.work.weixin.qq.com/document/path/99110

Feishu Sheets API: https://open.feishu.cn/document/server-docs/docs/sheets-v3/overview

Coze plugin guide: https://www.coze.cn/open/docs/guides/services

Part 3 – Implementation Details

Three‑Layer Framework

Interaction layer – Feishu group bots (smart & ordinary).

Conversion layer – Message forwarding to Coze and code‑logic encapsulation.

Execution layer – Coze automation:

Receive user message → large‑model intent detection → validation → update spreadsheet.

Scheduled triggers read the management table, compute due nodes, and push notifications.

Custom plugins retrieve coverage, static‑scan, and bug status via APIs.

Sample Workflow

# User mentions the bot
@ProjectManagementAssistant create task "Run static scan" owner @alice due 2023-04-20

# Bot extracts:
action: create task
title: Run static scan
owner: alice
due: 2023-04-20

# Validation passes → row added to Project Management Table

# Scheduler runs daily at 09:00
for each row where due = today:
    send "@alice, the task 'Run static scan' is due today."

# Plugin fetches static‑scan result
if scan fails:
    send "@alice, static scan failed. Please investigate."

Effectiveness

Input information is ingested automatically.

Test progress, bug summary, and branch coverage are generated and posted to the group.

Pre‑day reminders reduce missed static‑scan checks.

Automatic closure of TAPD items accelerates release cycles.

Future Work

The system is in early adoption; plans include extending the rule engine to additional workflow nodes and further reducing manual overhead.