Skip to content

execution-strategy.md

Latest commit

 

History

History
31 lines (22 loc) · 1.35 KB

execution-strategy.md

File metadata and controls

31 lines (22 loc) · 1.35 KB

Execution Strategy — Automatic Complexity Gating

Before starting any implementation task, silently evaluate its complexity and choose the right execution strategy.

Complexity Signals (count how many apply)

  • Touches 5+ files across 2+ modules
  • Requires new data flow (API → types → hooks → components)
  • Needs both UI components AND data layer changes
  • Requires schema/validation changes
  • Has dependencies between subtasks
  • Involves refactoring existing patterns across multiple files
  • Requires tests for new code
  • Touches auth, financial logic, or real-time data

Strategy Selection

  • 0-2 signals → Direct: Handle in main session. Use subagents for focused research only.
  • 3-5 signals → Parallel subagents: Spawn specialized subagents for independent work. Main session coordinates.
  • 6+ signals → Agent team: Spawn full team with dedicated teammates. Confirm with user first (higher token cost).

1M Context Recommendation

When a task requires reading 20+ files or spans 3+ modules, tell the user: "This task would benefit from 1M context. Consider restarting with claude --model opus[1m]."

Rules

  • Evaluate silently — don't narrate the assessment
  • Proceed immediately after choosing strategy
  • Only ask user confirmation for agent teams (higher cost)
  • For agent teams, suggest teammate split based on the specific task