Estimating Task Effort

Assign relative complexity estimates. Focus on comparison, not calendar time.

Quick Reference

• Input: Task list (from generate-tasks) or PRD requirements.
• Frameworks: Story points, t-shirt sizes, time ranges.
• Output: Each task annotated with estimate + confidence.
• Rule: Flag uncertainty - never fabricate precision.

HARD-GATE

DO NOT assign estimates you don't understand.
DO NOT default to "3 points" or "Medium" as a placeholder.
DO flag tasks needing further breakdown.
DO state confidence (high/medium/low) for every estimate.

Core Process

1. Receive - task list or PRD requirements.
2. Detect framework - ask user or infer from context (see Estimation Frameworks below).
3. Analyze each task - scope, unknowns, whether it needs further breakdown.
4. Assign - relative sizing, confidence level per task.
5. Flag - low-confidence tasks (recommend spike), oversized tasks (recommend split).
6. Output - annotated task list.

Estimation Frameworks

Mixing frameworks: Never mix frameworks within a single estimation table. If the user switches frameworks mid-session, re-estimate the full table.

Output Style

1. Framework - state which framework was used and why.
2. Estimation table - | ID | Task | Estimate | Confidence | Notes |
3. Flagged items - tasks needing spikes or splits, with reasoning.
4. Summary - total estimate, confidence distribution, callouts.
5. English only unless user requests otherwise.

Integration

Worked example: [assets/example-session.md](assets/example-session.md).