Skip to content

spec-implementer

The spec-implementer writes the code for a single assigned task. It is the core execution agent used by /spec-exec and spec-loop.sh.

Role

Implement the code for a task from tasks.md and ensure that code is fully wired into the running application before handing off to the tester.

Model

Sonnet. Implementation is well-defined work at this stage — the spec provides the architecture, the task provides acceptance criteria, and the codebase provides existing patterns to follow. Sonnet produces clean, accurate code quickly within these constraints.

When It Runs

  • /spec-exec — runs as the implementer in a single-iteration execution cycle
  • spec-loop.sh — invoked for each iteration of the automated implementation loop

What It Does

  1. Reads the assigned task from tasks.md along with requirements.md and design.md for context.
  2. Surveys the existing codebase to understand patterns and conventions before writing anything.
  3. Maps the wiring path — identifies exactly where and how the new code connects to existing code before writing it.
  4. Writes the implementation following existing codebase patterns.
  5. Wires the code in: adds imports, registers routes, updates navigation, connects API calls.
  6. Self-checks the wiring by reading modified files to confirm the chain is complete from entry point to new code.
  7. Updates tasks.md — sets Status: completed and Wired: yes (or n/a for infrastructure tasks).
  8. Reports to the Lead that the task is ready for testing.

Key Rules

  • Does not run tests — that is the tester's responsibility.
  • Does not mark Verified: yes — only the tester can set that field.
  • Does not review code quality — that is the reviewer's responsibility.
  • Wiring is non-negotiable. Code that exists but cannot be reached by a user is not considered complete. The implementer must confirm the full chain before marking a task done.
  • When receiving feedback from the debugger or Lead, fixes only the specific issues identified — no rewrites.