paidv0.1.0unverified
Spec Loop Pro — from vague idea to drift-checked feature, 4 skills
The spec-driven workflow as four composable Claude Code skills instead of a
framework install. No CLI to adopt, no lock-in: copy four folders, get the
spec-kit discipline with a benchmark ledger attached.
Quick start
Preview (free): skills/spec/ is included here — copy it into
.claude/skills/ and run /spec on your next feature idea.
Full bundle ($39): checkout link on the site → download all four skills +
exercises + the team-onboarding guide.
What you get
| Skill | Job | In preview? |
|---|---|---|
/spec | interview you, then write a testable one-page spec | ✅ free |
/spec-review | adversarial gate: attacks the spec for ambiguity, missing edges, untestable claims | paid |
/spec-to-tasks | decomposes an approved spec into ordered, verifiable tasks | paid |
/drift-check | diffs implementation against spec; reports drift before you merge | paid |
Why this beats the free frameworks
github/spec-kit (117k stars) and OpenSpec (58k) are excellent and free — and
they're frameworks: a CLI, a directory convention, a process to adopt. This
pack is the same discipline as four independent skills you can adopt one at a
time, and it's the only one that ships measured results per model release.
Proof
Unbenchmarked — v0. First benchmark: same feature built with and without the
loop, measuring rework commits and spec-to-implementation drift. Ledger lands
in BENCHMARKS.json before this pack goes on sale.
Changelog
- 0.1.0 (2026-07-04) — initial release, unverified. Not yet for sale.
One email when a pack ships or gets re-benchmarked on a new model. No noise.