Questionnaire

Every question feeds a specific formula variable. The skill asks only what's needed for the chosen path.

Quick Path (4 Questions)

#	Question	Feeds	Default
1	What needs to be done?	complexity, task_type	Inferred
2	What type of work?	task_type_multiplier	Inferred or asked
3	How many humans and agents?	num_humans, num_agents	1, 1
4	Automation maturity?	minutes_per_round	partial

Detailed Path (14 Questions)

All quick-path questions, plus:

#	Question	Feeds	Default
5	Risk factors?	risk_coefficient	1.3
6	Integration complexity?	integration_overhead	0.15
7	Domain familiarity?	domain_familiarity	1.0
8	Review depth?	review_minutes	standard
9	Human fix ratio?	human_fix_ratio	0.20
10	Confidence level?	confidence_multiplier	80% (1.4×)
11	Definition phase?	spread_multiplier	ready (1.0×)
12	Organization context?	org_overhead	solo-startup (1.0×)
13	Dependencies?	sequencing	none
14	Model & cost?	model_tier, show_cost	standard, false

Risk Factor Derivation (Detailed)

Instead of picking a number, the skill derives risk from specific concerns:

Factor	Increment
Unknown API surface / undocumented dependencies	+0.3
Ambiguous requirements	+0.2
New technology / unfamiliar domain	+0.3
External service dependencies	+0.2
Production data involved	+0.3

Base: 1.0. Add applicable increments. Cap at 2.5.

Batch Mode

For batch estimation, questions 2-12 are asked once as shared defaults. Then each task gets a confirmation table:

| # | Task           | Size | Type           | Risk | Deps | Adjust? |
|---|----------------|------|----------------|------|------|---------|
| 1 | Auth service   | M    | coding         | med  | —    |         |
| 2 | DB migration   | L    | data-migration | high | —    |         |

Users can override any task before estimation runs.

Input Formats (Batch)

The skill accepts tasks in any format:

• Plain list (one per line)
• CSV (task, complexity, type, notes)
• Markdown table
• JSON array
• Pasted from issue tracker

See the full questionnaire.md for complete details.