Power Tips

Last updated February 24, 2026

Once you’ve got the basics down, these patterns will make you significantly more effective.

Label taxonomy

Don’t invent labels on the fly — define a system upfront and stick to it:

## Label categories

Type:     #bug #feature #docs #chore #tech-debt
Area:     #frontend #backend #api #infra #design
Effort:   est:1h est:2h est:4h est:8h est:2d
Priority: priority:high priority:medium priority:low

Column conventions

Pick a column structure and use it everywhere. Consistency lets agents work across boards without extra context.

Standard development board:

# Backlog → # In Progress → # Review → # Done

Triage board:

# New → # Needs Info → # Accepted → # Won't Fix

Personal board:

# Today → # This Week → # Someday → # Done

Multi-board organization

As your project grows, use multiple focused boards instead of one giant board:

project/
├── sprint-current.md    # Active sprint tasks
├── backlog.md           # Prioritized feature backlog
├── bugs.md              # Bug tracker
├── decisions.md         # ADRs and decision log
└── retro-notes.md       # Sprint retrospective notes

Ask your agent to work across boards:

Move all completed items from sprint-current.md to an archive.
Pull the top 5 items from backlog.md into the new sprint.
Check bugs.md for any priority:high items and add them too.

Task decomposition

Big tasks are scary. Train your agent to break them down:

The task "Build user search" in sprint-12.md is too big (est:20h).
Break it down into subtasks of 2-4 hours each. Keep them in the
same column and add a label #search to group them. Mark the original
task as done and reference the subtasks.

This turns:

- [ ] Build user search @charlie #feature est:20h

Into:

- [x] Build user search (decomposed) @charlie #feature
- [ ] Set up Elasticsearch index @charlie #search #backend est:3h
- [ ] Build search API endpoint @charlie #search #backend est:4h
- [ ] Add search filters (date, type, status) @charlie #search #backend est:3h
- [ ] Build search UI component @alice #search #frontend est:4h
- [ ] Write search integration tests @bob #search #testing est:3h

Estimation calibration

Track your estimates vs. actuals to get better over time:

Look at the Done column of sprint-12.md. For each task, compare
the est: tag (estimated) with how long it actually took (if noted).
Which types of tasks do we consistently underestimate?

Smart due dates

Use due dates strategically — not everything needs one:

• priority:high tasks → always set a due: date
• priority:medium tasks → set due: only if there’s a real deadline
• priority:low tasks → skip due: entirely (they’ll get done when they get done)

Weekly review ritual

Every Friday, run this prompt:

Weekly review across all my boards:
1. What got Done this week? Celebrate the wins.
2. What's still In Progress? Is anything stuck?
3. Are there tasks older than 2 weeks in any column? Flag them.
4. What's the top 3 priority for next week?
Give me a summary I can share with the team.

Agent system prompts

If your AI agent supports system prompts or custom instructions, add context about your boards:

You have access to my AutoMD boards via MCP. Here's how they're organized:
- sprint-current.md: Active sprint tasks (columns: Backlog, In Progress, Review, Done)
- backlog.md: Product backlog ordered by priority
- bugs.md: Bug tracker (columns: Reported, Triaging, Fixing, Closed)

When I ask about "tasks" or "work", check sprint-current.md first.
When I report a bug, add it to bugs.md in the Reported column.
Always include @assignee and priority: when creating new tasks.

This front-loads context so every conversation starts smarter.