feat(design-sprint): add accelerated discovery sprint skill and docs

[ericthayer]

Summary

This pull request adds comprehensive support for the discovery phase in the product development workflow by introducing the design-sprint skill and related documentation. It clarifies the role of discovery in both command-driven (Claude) and intent-driven (OpenCode) environments, introduces a planned /discover command wrapper, and ensures that discovery is consistently mapped to the design-sprint process. The documentation and skill definitions have been updated to reflect these changes, providing clear guidance, process breakdowns, and facilitation prompts.

What changed

• Added new skill files:
  • skills/design-sprint/SKILL.md
  • skills/design-sprint/nudges.md
• Added lifecycle/intent routing in AGENTS.md:
  • Product discovery / pre-spec validation → design-sprint
  • DISCOVER → design-sprint
• Added future-wrapper placeholder issue template:
  • .github/ISSUE_TEMPLATE/discover-wrapper.md
• Updated docs (future-proofing, no command behavior change):
  • CLAUDE.md
  • docs/getting-started.md
  • docs/opencode-setup.md

Why this change

This fills the gap between ideation and formal specs:

• idea-refine (diverge/explore)
  → design-sprint (converge/validate)
  → spec-driven-development (formalize requirements)

The new skill uses a 5-phase flow:

Map & Target → Sketch → Decide → Prototype → Validate

with adaptive depth (Quick/Standard/Deep) and explicit outcomes (iterate/pivot/ship/kill).

Key Details:

1. New Skill: Design Sprint

• Added skills/design-sprint/SKILL.md, a detailed skill definition for design-sprint, outlining when and how to run a design sprint, adaptive depth selection, phase breakdowns, handoffs, verification, and red flags.
• Added skills/design-sprint/nudges.md, a facilitation prompt library for each phase of the sprint, including adaptive nudge selection and lens-based evaluation prompts.

2. Discovery Command Wrapper (Planned)

• Introduced .github/ISSUE_TEMPLATE/discover-wrapper.md describing the plan for a /discover (and optional /sprint) command wrapper that invokes the design-sprint skill, including requirements, goals, and acceptance criteria.
• Updated CLAUDE.md and docs/getting-started.md to mention the planned /discover command and clarify that it will serve as a convenience wrapper for design-sprint. [1] [2]

3. Documentation Updates: Discovery in the Workflow

• Updated AGENTS.md and docs/opencode-setup.md to explicitly map the discovery phase and high-uncertainty situations to the design-sprint skill, both in intent mapping and lifecycle descriptions. [1] [2] [3] [4]
• Clarified that command wrappers like /discover are optional UX for Claude and not required in OpenCode, which remains intent-driven.
• Added discovery phase to workflow diagrams and phase tables in CLAUDE.md and docs/getting-started.md. [1] [2]

4. User Experience and Guidance

• Updated command tables and instructions in docs/getting-started.md to include /discover as a planned command and reinforce that skills are the source of truth, not commands. [1] [2]

These changes collectively provide a clear, robust, and documented approach to running discovery sprints, ensuring both agent and user workflows are aligned and well-supported.

• [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13]

Verification

• New files pass diagnostics:
  • skills/design-sprint/SKILL.md ✅
  • skills/design-sprint/nudges.md ✅
  • .github/ISSUE_TEMPLATE/discover-wrapper.md ✅
• File sizes:
  • SKILL.md: 212 lines
  • nudges.md: 168 lines
• Packaged archive verified to include expected files.

Rollout notes

• Additive and non-breaking.
• Discovery is available now via skill routing (AGENTS.md).
• /discover (optional /sprint) is documented as planned wrapper UX only (not implemented in this PR).
• OpenCode remains intent-driven (no slash-command requirement).

Risks

• New /discover wrapper is documented as planned but intentionally not implemented in this PR (skill-first behavior remains source of truth).
• Existing markdown-lint warnings in legacy docs (AGENTS.md and other repo docs) are pre-existing and not introduced as behavioral changes.