spec-driven dev kit

"complete spec driven dev kit; code is literally just artifact." the HumanLayer talk describes exactly this: a Research → Plan → Implement pipeline with context management for coding agents. the philosophy: a bad line of research (misunderstanding how the system works) can cause thousands of bad lines of code, so you invest most time in specifying the right problem and understanding the system before launching the coding agent.

key components: research prompts that output file names and line numbers so agents know exactly where to look, planning prompts that specify every change with files and snippets, implement prompts that maintain context under 40% utilization, intentional human review at each phase. the research file is a lot easier to review than a 2000-line PR. the workflow keeps mental alignment across the team.

connects to the hard software component docs writer for the codebase understanding layer, and to the overnight app grinder which would benefit from spec-first architecture.

spreadsheet evaluation: tech depth 8/10, labeled [DO THIS] — "you know coding agents well, could build tooling others actually use." blue ocean analysis: RED OCEAN positioning with medium 2-month feasibility — the market is competitive but the approach is strong. listed as one of 5 [DO THIS] tier projects.

timeline

• [2026-04-10] captured from google sheets — [DO THIS] tier, RED OCEAN but strong technical fit