productivity

# Productivity

Umbrella skill for personal and team productivity: task decomposition, daily/weekly planning, meeting optimization, status updates, goal setting, and focus management. Each mode loads its own reference files on demand.

---

## Mode Detection

Classify into one mode before proceeding.

| Mode | Signal Phrases | Reference |
|------|---------------|-----------|
| **TASK** | add task, prioritize tasks, task list, what's on my plate, decompose work, batch tasks | `references/task-management.md` |
| **PLAN** | daily plan, plan my day, time blocks, plan my week, energy mapping | `references/daily-weekly-planning.md` |
| **MEETING** | meeting agenda, optimize meeting, meeting audit, cancel this meeting, async alternative | `references/meeting-optimization.md` |
| **STATUS** | status update, standup, weekly update, stakeholder update, progress report | `references/status-updates.md` |
| **REVIEW** | weekly review, retro, retrospective, reflect on week, monthly review | `references/daily-weekly-planning.md` |
| **GOAL** | set goals, OKRs, quarterly goals, goal progress, key results | `references/daily-weekly-planning.md` |

If the request spans modes, pick the primary mode and note the secondary.

---

## Workflow by Mode

### TASK Mode

**Load**: `references/task-management.md`, `references/llm-productivity-failure-modes.md`

1. **Capture** — Accept tasks in any format: freeform text, bullet lists, pasted meeting notes, vague intentions. Extract actionable items.
2. **Decompose** — Apply vertical slicing (because horizontal slices create work that cannot ship independently):

   | Slice Quality | Example |
   |---------------|---------|
   | Good (vertical) | "User can upload a CSV and see a preview" — shippable alone |
   | Weak (horizontal) | "Build the upload API" — requires the UI to deliver value |

3. **Estimate** — Assign time estimates using the 1/2/4-hour bucketing system (because finer granularity creates false precision, coarser loses planning value). Tasks over 4 hours need decomposition.
4. **Prioritize** — Apply the appropriate framework based on context:

   | Context | Framework | Why |
   |---------|-----------|-----|
   | Personal daily work | Eisenhower (urgent/important) | Fast, intuitive, separates reactive from proactive |
   | Backlog with many items | ICE (Impact/Confidence/Ease) | Quantitative ranking without heavy data requirements |
   | Team sprint planning | Weighted scoring against goals | Defensible, transparent to stakeholders |

5. **Organize** — Group by context (because context-switching between unrelated tasks costs 15-25 minutes per switch). Batch similar work: all emails together, all code reviews together, all writing together.

**Gate**: Every task has an action verb, a completion condition, and a time estimate. Vague items like "think about marketing" get reframed as "Draft 3 marketing channel options with pros/cons (2h)."

### PLAN Mode

**Load**: `references/daily-weekly-planning.md`, `references/llm-productivity-failure-modes.md`

1. **Gather constraints** — Ask for (conversationally, not as a wall of questions):
   - Calendar commitments for the day/week
   - Hard deadlines
   - Energy level and known energy patterns (because matching task difficulty to energy state increases completion rates)
   - Carryover from yesterday

2. **Select top priorities** — Identify the Top 3 outcomes for the day (because more than 3 priorities means zero priorities). Apply the "if only these 3 things got done, would today feel successful?" test.

3. **Build time blocks** — Map tasks to calendar slots:

   | Block Type | When to Schedule | Duration |
   |------------|-----------------|----------|
   | Deep work (creation, analysis) | Peak energy hours (usually morning) | 90-120 min |
   | Reactive work (email, Slack, reviews) | Low energy hours (usually post-lunch) | 30-60 min batches |
   | Admin/maintenance | End of day | 30 min |
   | Buffer | Between blocks | 15 min minimum |

4. **Identify conflicts** — Flag when calendar meetings fragment deep work blocks. Surface the cost: "You have 3 meetings between 9-12, leaving zero uninterrupted blocks during your peak hours."

5. **Generate plan** — Output a concrete, time-blocked plan with the Top 3 outcomes highlighted.

**Gate**: Plan accounts for actual calendar (not aspirational free time). Deep work blocks are at least 90 minutes. Buffers exist between blocks. Total planned work does not exceed available hours minus 20% (because unplanned work always appears).

### MEETING Mode

**Load**: `references/meeting-optimization.md`, `references/llm-productivity-failure-modes.md`

1. **Determine operation**:

   | Operation | What to Do |
   |-----------|-----------|
   | Audit existing meeting | Apply the 5P framework: Purpose, Participants, Preparation, Process, Payoff |
   | Design new agenda | Build outcome-driven agenda with time allocations and decision types |
   | Convert to async | Draft async alternative with decision framework and deadline |
   | Optimize recurring meeting | Analyze frequency, attendance, decision output vs time spent |

2. **For audits** — Calculate meeting cost (participants x hourly rate x duration x frequency). Surface the number because most people underestimate it. A weekly 1-hour meeting with 8 people at $75/hr costs $31,200/year.

3. **For agendas** — Every agenda item gets:
   - **Type**: Decision, Discussion, Information, or Brainstorm (because different types need different facilitation)
   - **Owner**: Who presents/facilitates this item
   - **Time**: Allocated minutes
   - **Pre-read**: What participants should review before the meeting
   - **Outcome**: What "done" looks like for this item

4. **For async conversion** — Apply the async-first decision tree:
   - Can this be a document with comments? Do that instead.
   - Does this need real-time debate? Keep the meeting, shorten it.
   - Does this need a decision from one person? Send them a 1-page memo with a deadline.