Spec Intake Wizard

SDD Spec Intake Wizard Skill

Stateless atomic operation to engage the user in a high-fidelity, interactive intake flow. It prompts the user for essential system parameters, recommends baseline defaults to expedite answers, validates constraints in real time, and outputs a normalized JSON metadata structure that feeds downstream specification generators (spec-generator) and planners (task-planner).

Role & Goal

• Role: Technical Product Manager / Interactive Intake Engine.
• Goal: Gather high-fidelity, unambiguous requirements from the user with zero friction.

Steps

Step 1: initialize_wizard_schema

Initialize the interactive session with a dynamic schema based on the targeted domain or feature category:

• Define the standard template of required fields:
  • feature_name: String (required, 2-100 characters)
  • target_domain: String (e.g. "infra", "finance", "dev-workflows", "ops", "research")
  • high_level_goal: String (required, summary of intent)
  • key_constraints: List of strings (e.g. "latency < 200ms", "no external databases")
  • expected_inputs: List of fields and types
  • expected_outputs: List of fields and types
• Populate a set of smart defaults and "recommended" answers to enable a single-click or empty-input acceptance.
• Output parameters:
  • session_schema: The formulated validation schema and prompt rules.

Step 2: run_clarification_loop [depends_on: initialize_wizard_schema]

Interact with the user to collect parameters and resolve the top ambiguity gaps:

• Present the structured questions sequentially or as a single clear interactive prompt.
• For each option, clearly present the recommended default (e.g., "[Default: YES]" or "[Default: patch]").
• Allow the user to input custom values, accept defaults by hitting Enter, or skip the remaining loop by typing "skip".
• Gracefully handle empty or invalid entries with supportive inline validation help messages.
• Output parameters:
  • raw_answers: Key-value map of collected user responses.

Step 3: compile_validated_intake [depends_on: run_clarification_loop]

Execute final validation, reconcile inputs against the target default schema, and produce a unified JSON configuration:

• Merge raw_answers with session_schema defaults for any omitted or skipped fields.
• Assert strict structural validation constraints (e.g., ensure mandatory keys are present and data types match).
• Parse constraints and requirements into a normalized data model structure:
  • metadata: { created_at: String, user: String, skip_flag: Boolean }
  • spec_intake: { name: String, domain: String, goal: String, constraints: List, data_flow: Object }
• Output parameters:
  • status: "SUCCESS" or "FAILED"
  • payload: Standardized JSON representation of the user requirements.
  • summary_markdown: A clean, structured markdown table representing the chosen configuration parameters.