AgentSkillsCN

Blossom

绽放

SKILL.md

Blossom Skill

Spike-driven exploration workflow for discovering solutions before committing to implementation.

Usage

code
/blossom <goal>

Workflow

Phase 1: Understand

Gather context about the goal:

  • What modules/files are involved?
  • What interfaces/protocols constrain the solution?
  • What tests already exist?

Phase 2: Spike

Create a minimal throwaway spike to test the approach:

  • Write the smallest possible code that validates the idea
  • Run it (tests, REPL, or script)
  • Capture what you learned

Phase 3: Evaluate

Assess the spike:

  • Does it work? What broke?
  • Does it fit the existing architecture patterns?
  • What edge cases exist?
  • Is the approach worth pursuing?

Phase 4: Propose

Present findings:

  • What worked, what didn't
  • Recommended approach with rationale
  • Estimated scope (files to change, tests to write)
  • Any risks or open questions

Spike Areas for This Project

  • Event Store backends: New storage backend implementations
  • Bus integrations: New event bus implementations (message brokers)
  • Projection patterns: New projection types or processing strategies
  • Serialization: Event versioning, schema evolution
  • Multi-tenancy: Tenant isolation strategies
  • Observability: Tracing, metrics, logging patterns
  • Migration tooling: Event store migration utilities

Rules

  • Spikes are throwaway -- never commit spike code directly
  • Always run existing tests after spiking to verify nothing broke
  • If the spike touches a protocol/interface, check all implementations