Run Technical Discovery

Purpose

Help teams run a time-boxed technical discovery process that surfaces feasibility constraints, explores architectural options, and resolves engineering unknowns — before committing to delivery.

Skill type

Conceptual skill

Use this skill when

• A product initiative has significant technical uncertainty that could invalidate the plan
• Engineering can't provide reliable estimates without exploration time
• A novel approach (new AI capability, new integration, new platform feature) needs to be validated
• An architectural decision needs structured exploration before a path is chosen

Do not use this skill when

• Technical complexity is low and discovery isn't needed
• The goal is sprint planning for known work (use plan-delivery-collaboration)

Required inputs

• Product initiative with technical uncertainty
• Key technical questions that must be resolved

Optional inputs

• Existing system architecture
• Technical constraints (data residency, security, performance)
• Engineering team input on unknowns
• Desired time box for discovery

Upstream context

Works best when:

• Product initiative is defined (problem and goals)
• Engineering has flagged unknowns

Downstream handoff

Output can feed:

• write-requirements-prd (discovery findings → requirements)
• plan-delivery-collaboration (discovery outputs enable reliable estimates)
• manage-technical-debt-tradeoffs (discovery surfaces debt implications)

Instructions

1. Define the key technical questions to resolve in discovery.
2. Agree on the time box (typically 1–2 sprints).
3. Define the definition of done: what does discovery need to produce to unblock planning?
4. Identify spike tasks or prototype experiments needed to answer each question.
5. Define how findings will be documented and shared.
6. Plan a discovery readout to align product and engineering on implications.

Output

Provide:

• Technical questions to resolve
• Discovery time box
• Definition of done
• Spike tasks or experiments per question
• Documentation and sharing plan
• Discovery readout format
• Decision points that discovery will unlock

Risks / caveats

• Open-ended discovery without time boxing turns into research that delays delivery indefinitely
• Discovery output must be in a form product can use — not just internal engineering notes
• "We need more discovery" is sometimes avoidance — distinguish genuine unknowns from lack of confidence