Skip to content

Latest commit

 

History

History
177 lines (122 loc) · 4.06 KB

File metadata and controls

177 lines (122 loc) · 4.06 KB

Dogfooding Guide

This guide is for using Taskdeck daily while it remains capture-first, board-centered, and review-first.

Primary goal:

  • reduce maintenance overhead by capturing quickly
  • turn capture into reviewed proposals, not silent automation
  • find product-legibility friction before adding more surface breadth

Best Current Dogfooding Shape

Taskdeck is strongest right now for:

  • solo developers
  • builders/operators
  • anyone who wants safer AI-assisted work intake without silent board mutation

Use it as:

  • capture
  • triage
  • review
  • board execution

Do not expect it yet to be:

  • a polished novice-first app shell
  • a broad autonomous project manager
  • a replacement for every planning/reporting surface

Recommended Workspace Setup

Keep the workspace intentionally small at first.

Recommended boards:

  1. active product board
  2. backburner or ideas board
  3. demo or experiments board

Recommended columns:

  • Backlog
  • Ready
  • In Progress
  • Review
  • Done

Recommended starter labels:

  • priority-high
  • bug
  • tech-debt
  • blocked
  • demo

Reason:

  • the goal is to test the core loop, not taxonomy complexity

Current MVP Loop

  1. Capture into Inbox
  • use Inbox for notes, bugs, follow-ups, and loose ends
  • do not pre-sort during capture; save it first
  1. Triage into proposals
  • open Inbox daily or twice daily
  • ignore noise and triage the rest
  • let triage create proposals asynchronously
  1. Review proposals
  • open Automations -> Proposals
  • validate operations before approval
  • execute only after review
  1. Work the board
  • keep the board as the visible work surface
  • move cards honestly
  • use comments to preserve reasoning

Daily Rhythm

Morning (5 to 10 min):

  • triage Inbox
  • review pending proposals
  • pick 1 to 3 cards for active work

During work:

  • capture follow-ups immediately
  • do not reorganize while capturing
  • use comments or mentions when context would otherwise get lost

End of day (3 to 5 min):

  • move completed work forward
  • capture unfinished work into Inbox
  • avoid ending the day with critical context only in your head

Weekly (15 to 30 min):

  • archive or move stale work
  • check which board is truly active
  • review friction notes from the week

Automation Guardrails

Use automation for:

  • repetitive edits
  • templated card creation
  • low-risk board hygiene

Avoid automation for:

  • high-context decisions without human review
  • broad changes you cannot quickly inspect

Queue guidance:

  • treat Queue as advanced
  • keep requestType as instruction
  • use board context when available
  • prefer one clear instruction per request

Chat guidance:

  • prefer board-scoped sessions
  • use it when you want conversational planning or board bootstrap
  • keep proposal review as the actual mutation gate

What To Log

Keep a lightweight friction log for each meaningful issue:

  • objective
  • expected behavior
  • actual behavior
  • minimal repro
  • impacted workflow

High-value friction themes:

  • poor discoverability
  • empty states with no next-step guidance
  • flows that still feel too internal or system-shaped
  • steps where you asked "what should I click next?"
  • automation failures with unreadable recovery guidance

Healthy Signs

  • you capture more often because it is cheap
  • Inbox reaches zero or near-zero regularly
  • proposals feel safe rather than mysterious
  • the board becomes the place where work gets finished
  • you need fewer external notes to remember follow-ups

Unhealthy Signs

  • you avoid Inbox because triage feels annoying
  • you bypass review because the value is unclear
  • you use Queue for ordinary work
  • you still need internal IDs or hidden knowledge too often
  • you keep context in text files because Taskdeck feels slower than capture elsewhere

Practical Success Criteria

Track these sanity checks:

  • Inbox reaches zero or near-zero most days
  • proposals execute without manual DB intervention
  • a new user could reach first useful card in under 2 minutes with guidance
  • failed automations return readable, actionable errors
  • the board remains the system of visible work, not just a side effect of automation