BMAD Ate My Month: When Spec-Driven Development Is Too Much

Description and use

An autonomous coder produced large PRs causing costly reviews. Switching to a compact Spec Kit with pinned SHA and ~300-line work limits reduced token usage and accelerated releases.

Problem

Autonomous coding agents produced large, poorly planned pull requests that caused costly review churn and stalled releases.

Outcome

By replacing BMAD’s encyclopedic planning with a concise, pinned-spec approach, the author reduced review churn, token usage, and restored release velocity.

Workflow

• 1. Create a user story issue with a pinned Spec Kit reference (full SHA).
• 2. If a story has >12 tasks, split it before filing.
• 3. For multiple stories, create a parent epic issue with one child issue per story and an integration branch.
• 4. Run the autopilot skill to generate code against the pinned spec.
• 5. Review the PR using a BLOCKING/SHOULD/NIT rubric tied to the spec.
• 6. Merge if all checks pass; otherwise iterate.
• 7. Keep the spec and issue small to avoid large diffs and long review cycles.

Evidence

I did not adopt another framework. I took the one thing from the spec-driven world that had earned its keep, Spec Kit's way of expressing intent, and forced it into add-autopilot-issue

Source

Every story issue carries a pinned spec reference: Spec: owner/repo@sha:path#user-story-N

Source