linyuhang
/
PMFlow
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
PMFlow/docs/superpowers/plans/2026-04-02-feishu-web-full-...

3.3 KiB
Raw Blame History

Feishu Web Full Flow Implementation Plan

For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (- [ ]) syntax for tracking.

Goal: Add a real Feishu web automation script that drives the PMFlow draft-confirmation flow end to end, and make project creation auto-assign generated tasks through the skill graph.

Architecture: Keep the existing draft_pending_confirm PMFlow flow as the orchestration backbone. The backend change assigns owners during AI Steward project creation by resolving skill tags and calling the existing personnel recommender. The browser change uses Playwright with a persistent Chrome profile to send real Feishu web messages, then verifies downstream results through the database.

Tech Stack: FastAPI, SQLAlchemy, React/Vite workspace tooling, Playwright, Node.js, PostgreSQL

Task 1: Lock task auto-assignment with a failing backend test

Files:

  • Modify: backend/tests/test_ai_steward_project_from_text_contract.py

  • Add a unit test for _normalize_project_creation_draft that expects a generated task draft to include assignee_id and structured recommendation metadata when the skill recommender returns a top match.

  • Run the focused pytest target and confirm it fails for the missing assignment behavior.

Task 2: Implement skill-graph-based auto-assignment in AI Steward

Files:

  • Modify: backend/app/services/ai_steward_service/shared.py

  • Modify: backend/app/services/ai_steward_service/mixin_creation_normalize_persist.py

  • Import the personnel recommender into the AI Steward service layer.

  • Add a helper that requests recommendations from the skill graph and normalizes the top candidate into assignee_id plus ai_recommended_assignees.

  • Thread the recommendation result into build_task_create_draft.

  • Re-run the focused pytest target and confirm it passes.

Task 3: Add Playwright Feishu web automation scaffolding

Files:

  • Modify: frontend/package.json

  • Add: frontend/playwright.config.ts

  • Add: frontend/e2e/feishu-project-full-flow.spec.ts

  • Add Playwright dev dependencies and test scripts.

  • Add a config that supports a persistent Chrome session, local artifacts, and env-driven runtime settings.

  • Write the Feishu web E2E that sends a unique project-creation message, revises the draft for 3-5 rounds with explicit work_item targeting, confirms the draft, and validates database side effects.

Task 4: Update project docs

Files:

  • Modify: CLAUDE.md

  • Document the new frontend/playwright.config.ts and frontend/e2e/ test directory in the structure sections.

  • Document the AI Steward task auto-assignment behavior and the new Feishu web E2E coverage where relevant.

Task 5: Verify and iterate until green

Files:

  • No code changes expected, but rerun as needed

  • Run focused backend pytest for the new auto-assignment behavior.

  • Install frontend dependencies and Playwright.

  • Run the Feishu web E2E and fix selectors, environment assumptions, or backend gaps until it passes in the current environment.

  • Run python -m py_compile on changed backend modules and npm run build / tsc --noEmit in frontend if dependency/setup state permits.