chore: auto-commit after execute-task

GSD-Unit: M001/S06/T01
This commit is contained in:
2026-03-27 08:54:34 +01:00
parent 7a44ebbd77
commit bdc47dde7f
31 changed files with 2017 additions and 966 deletions
+13 -130
View File
@@ -1,132 +1,15 @@
# M001: Gmail and draft quality loop
# M001: M001: Gmail and draft quality loop
**Vision:** Turn the existing job tracker into a daily-use personal job-search workspace where Gmail import and AI drafting are strong enough to trust, while preserving manual control over all real-world sending and applying.
## Vision
Turn the existing job tracker into a daily-use personal job-search workspace where Gmail import and AI drafting are strong enough to trust, while preserving manual control over all real-world sending and applying.
## Success Criteria
- User can import a job found elsewhere, generate a tailored CV and cover-letter package that feels specific enough to start from, and save/edit that package inside the job workspace.
- User can connect Gmail, import the right message or full thread into a job with less cleanup than before, and see the imported correspondence reflected in that jobs timeline/workspace.
- After a thread is linked to a job, later inbound messages and user-sent Gmail replies appear on that job automatically without requiring the user to re-import the thread manually.
- User can generate a follow-up or reply draft grounded in the imported correspondence and saved job/application context.
- The daily loop of job table → follow-up/dashboard → individual job workspace feels coherent and actionable for an individual user.
- No part of the milestone auto-sends email or auto-applies to jobs.
## Key Risks / Unknowns
- Gmail import may still feel unreliable if matching, whole-thread import, and linked-thread refresh clarity do not improve enough to reduce manual cleanup.
- AI draft quality may still feel generic even though the draft surfaces already exist.
- The workflow may remain fragmented if table/dashboard/job-detail changes do not land as one coherent loop.
- Real value may depend on live Gmail + AI + persisted job context wiring, not isolated endpoint improvements.
- Automatic Gmail thread refresh may be harder than one-time import if Gmail history/state handling is brittle or expensive.
## Proof Strategy
- Gmail import reliability and trust → retire in S01 by proving a user can connect Gmail, review likely messages or threads for a job, import the right correspondence into that job, and keep a linked thread current as new replies arrive.
- AI draft quality and usefulness → retire in S02 by proving imported job context plus profile/CV context produce tailored drafts the user can edit and save as actual working material.
- Reply/follow-up assistance grounded in real context → retire in S03 by proving imported correspondence and saved draft state feed a useful reply/follow-up drafting flow.
- Workflow coherence across daily surfaces → retire in S04 and S05 by proving the table/dashboard/job workspace work as one control loop and by re-checking the whole loop end-to-end.
## Verification Classes
- Contract verification: backend/frontend tests, artifact checks for new endpoints and UI flows, persisted state checks, import/draft wiring verification
- Integration verification: real Gmail OAuth/import plus live AI-service-backed draft generation exercised through the app
- Operational verification: repeated use of the workflow across auth/config/service boundaries without dangerous outbound automation
- UAT / human verification: whether Gmail import feels trustworthy, whether linked threads stay current automatically enough for real use, and whether drafts feel strong enough to start from
## Milestone Definition of Done
This milestone is complete only when all are true:
- all slice deliverables are complete
- Gmail import, linked-thread continuity, correspondence state, and draft-generation surfaces are actually wired together
- the real browser entrypoint exists and is exercised through the table/dashboard/job loop
- success criteria are re-checked against live behavior, not just artifact presence
- final integrated acceptance scenarios pass
## Requirement Coverage
- Covers: R001, R002, R003, R004, R005, R006, R007, R008, R010
- Partially covers: R009
- Leaves for later: R011, R012, R013
- Orphan risks: none
## Slices
- [x] **S01: Smarter Gmail import and matching** `risk:high` `depends:[]`
> After this: User can connect Gmail, review likely messages or threads for a job, import a message or full thread, and trust linked Gmail threads to stay current on that job without manual re-import.
- [x] **S02: Stronger AI application package drafting** `risk:high` `depends:[S01]`
> After this: From an imported job plus profile/CV context, the app generates materially better tailored CV and cover-letter drafts that feel specific and usable.
- [x] **S03: Reply and follow-up drafting from real thread context** `risk:medium` `depends:[S01,S02]`
> After this: Inside a job, the user can generate follow-up and reply drafts grounded in imported and automatically refreshed correspondence plus saved application context, then edit them before sending manually.
- [x] **S04: Daily control loop surfaces** `risk:medium` `depends:[S01,S03]`
> After this: The job table works as the primary overview and the follow-up/dashboard surfaces clearly show what needs attention next for an individual user.
- [x] **S05: End-to-end trust and workflow polish** `risk:low` `depends:[S01,S02,S03,S04]`
> After this: The full loop works cleanly in a real environment: import job → generate package → apply externally → import/update correspondence automatically from linked Gmail threads → draft follow-up/reply → track progress confidently.
- [ ] **S06: Live environment stabilization and integrated acceptance rerun** `risk:high` `depends:[S05]`
> After this: The real M001 environment runs without the current backend/frontend CORS/runtime blockage, and the full `/jobs` → workspace → Gmail continuity → follow-up → dashboard/reminders loop is re-checked live with recorded acceptance results.
- [ ] **S07: Daily-loop UAT artifact closure** `risk:medium` `depends:[S06]`
> After this: The overview-surface/browser validation is captured as a real executed UAT artifact instead of a placeholder, proving the same job behaves coherently across table, dashboard, reminders, and workspace entry.
## Boundary Map
### S01 → S02
Produces:
- improved Gmail import workflow in `job-tracker-ui/src/components/Correspondence.tsx` that yields job-linked imported correspondence with clearer message/thread selection behavior
- stronger backend correspondence import surface in `JobTrackerApi/Controllers/GmailController.cs` and related persistence so imported messages reliably attach to `JobApplication` records
- linked-thread metadata and refresh behavior stable enough that later draft-generation flows can consume correspondence as trusted context rather than a stale snapshot
Consumes:
- nothing (first slice)
### S01 → S03
Produces:
- imported correspondence records tied to a specific job and available through the existing per-job correspondence/timeline surfaces
- message/thread metadata good enough for later reply/follow-up draft context assembly
- a verified Gmail connection/import path that downstream slices can rely on, including ongoing thread refresh after first import
Consumes:
- nothing (first slice)
### S02 → S03
Produces:
- improved application package generation via `JobTrackerApi/Controllers/JobApplicationsController.cs` returning stronger tailored CV and cover-letter outputs tied to a job
- persisted draft state in the job workspace so later follow-up/reply flows can reuse saved application context
- clearer frontend editing/saving behavior in `job-tracker-ui/src/components/JobDetailsDialog.tsx`
Consumes from S01:
- imported and auto-refreshed correspondence plus job-linked message context
### S03 → S04
Produces:
- reply/follow-up draft flow grounded in job + correspondence + saved application package context
- explicit manual-send boundary in the job workspace UI and backend behavior
- job-level indicators that a follow-up/reply is ready, missing context, or needs action
Consumes from S01:
- Gmail-imported and automatically refreshed correspondence context
Consumes from S02:
- saved tailored CV / cover-letter / application package context
### S04 → S05
Produces:
- table/dashboard surfaces that summarize readiness, follow-up urgency, and next actions for individual users
- clearer navigation hierarchy across table, follow-up/dashboard, and individual job workspace
- stable daily-use control loop to validate in final integration
Consumes from S01:
- correspondence state, Gmail import outcomes, and linked-thread refresh outcomes
Consumes from S03:
- follow-up/reply drafting signals and job-level action state
## Slice Overview
| ID | Slice | Risk | Depends | Done | After this |
|----|-------|------|---------|------|------------|
| S01 | Smarter Gmail import and matching | high | — | ✅ | User can connect Gmail, review likely messages or threads for a job, import a message or full thread, and trust linked Gmail threads to stay current on that job without manual re-import. |
| S02 | Stronger AI application package drafting | high | S01 | ✅ | From an imported job plus profile/CV context, the app generates materially better tailored CV and cover-letter drafts that feel specific and usable. |
| S03 | Reply and follow-up drafting from real thread context | medium | S01, S02 | ✅ | Inside a job, the user can generate follow-up and reply drafts grounded in imported and automatically refreshed correspondence plus saved application context, then edit them before sending manually. |
| S04 | Daily control loop surfaces | medium | S01, S03 | ✅ | The job table works as the primary overview and the follow-up/dashboard surfaces clearly show what needs attention next for an individual user. |
| S05 | End-to-end trust and workflow polish | low | S01, S02, S03, S04 | ✅ | The full loop works cleanly in a real environment: import job → generate package → apply externally → import/update correspondence automatically from linked Gmail threads → draft follow-up/reply → track progress confidently. |
| S06 | Live environment stabilization and integrated acceptance rerun | high | S05 | ⬜ | The real M001 environment runs without the current backend/frontend CORS/runtime blockage, and the full `/jobs` → workspace → Gmail continuity → follow-up → dashboard/reminders loop is re-checked live with recorded acceptance results. |
| S07 | Daily-loop UAT artifact closure | medium | S06 | ⬜ | The overview-surface/browser validation is captured as a real executed UAT artifact instead of a placeholder, proving the same job behaves coherently across table, dashboard, reminders, and workspace entry. |