Changelog

Every notable change to the skills is recorded here. The format follows Keep a Changelog, and the project uses Semantic Versioning. While the version is 0.x, a minor version can contain breaking changes and a patch version does not.

Unreleased

0.2.0 - 2026-11-05

Breaking changes are marked Breaking. A gate’s status values changed, so update anything that reads a **Status:** line.

Added

• Claude Code plugin sdd in the georgeracu marketplace: claude plugin marketplace add georgeracu/sdd-skills, then claude plugin install sdd@georgeracu.
• sdd:setup, which copies the scaffold into a project, records the build and test commands in AGENTS.md and reports the optional inputs it found.
• Documentation site at https://georgeracu.github.io/sdd-skills/, with a page for each skill, the worked example, this changelog and an RSS feed.
• ROADMAP.md and CONTRIBUTING.md.
• spec-implementation-audit checks every correctness property in design.md against the code that enforces it and the test that exercises it.

Changed

• Breaking: the scaffold moved from scaffold/ to skills/setup/scaffold/, so every install route carries it. Run /sdd:setup instead of copying it by hand.
• The five review skills share one verdict model: a single **Status:** line, a status that follows the highest severity, and a second blocking run that asks the user how to continue (#24).
• Breaking: spec-design-review renames the status APPROVED WITH MINOR ISSUES to APPROVED WITH ISSUES. Every finding names its source, identifier drift is MAJOR, and an unsupported latency criterion, a data exposure or an auth bypass is CRITICAL (#25).
• spec-task-review: a trace from a test task alone does not cover a criterion, sequencing names the artefact and the tasks that create and use it, and discovery.md is deleted on both passing verdicts (#26).
• Breaking: spec-qa-review uses the **Status:** line for PASSED or FAILED, and moves the lifecycle value (Draft, Active, Complete) to a **Stage:** line. Unit coverage is gated on line and branch coverage at 85% each, and the severity scale is CRITICAL, MAJOR and MINOR (#27).
• Breaking: spec-implementation-audit returns FAILED on any CRITICAL or MAJOR finding, where the earlier version did not say which severities fail, and stops after writing a FAILED report instead of fixing the findings. Evidence blocks hold the command and its verbatim output, and only a ticked box completes a task (#28).
• Breaking: spec-maintenance counts a work item as implemented only when its implementation-audit.md says **Status:** PASSED, where any audit file counted before, and the current item’s own audit no longer counts. A run with no such peers writes no report. Annotations take one shape, and the drift warning counts one impact per bullet (#29).
• The README replaces “no promise of support” with a stability promise.

0.1.3 - 2026-10-20

Added

• sdd:spec-discovery, which agrees the problem, scope and approach with the user and writes discovery.md. It replaces the dependency on superpowers’ brainstorming skill for step 1.
• scaffold/, the project/work-items/ layout with process documents and templates, ready to copy into a project.
• The spec 38 worked example in examples/, with a README that maps each artefact to the skill that produced it.
• Eval packs for the five gate skills.

Changed

• The skills are stack-neutral: they read the build, test and type-generation commands from AGENTS.md instead of assuming Gradle or npm.

Fixed

• spec-generator creates project/work-items/ when it is absent.
• spec-discovery skips questions the user has already answered.

0.1.2 - 2026-09-07

Changed

• spec-generator: the bugfix template enumerates the neighbouring behaviours that must not change (#14), and the path after a BLOCKED design review drafts the fixes (#13).
• The spec-generator eval pack grows to nine cases with per-case criteria.

0.1.1 - 2026-09-07

Added

• The first eval pack for spec-generator.

0.1.0 - 2026-09-07

Added

• Seven skills extracted from a private project: spec-workflow, spec-generator, spec-design-review, spec-task-review, spec-qa-review, spec-implementation-audit and spec-maintenance.