Appearance
Release Notes
This page tracks release-facing changes at a higher level than the raw changelog.
0.3.3 — Discovery Interview UX patch
Changed
- Stronger interview behavior:
spdd-discoverynow treats the user's initial prompt as a starting hypothesis and asks focused questions before writing a Discovery Brief. generate nowhandoff: users can explicitly saygenerate nowwhen the context is good enough to write the brief.- REASONS-informed checklist: discovery now uses a concise coverage checklist for Requirements, Entities, Approach, Structure, Operations, Norms, and Safeguards to guide the next question.
- Adaptive questioning: after each relevant answer, the skill re-evaluates the best next question from the current context instead of following a fixed script.
- Prompt sync: the discovery SPDD prompt now matches the implemented local command wrapper and interview behavior.
Notes
- This patch does not add automatic orchestrator discovery routing or
with-discoverycontrols.
0.3.2 — Discovery Interview and local command wrappers
Added
spdd-discoveryskill: manual Discovery Interview skill for rough ideas, noisy notes, pasted context, or referenced files that need a durable Discovery Brief before planning, story splitting, or technical analysis.- Discovery Brief artifacts: completed discovery runs write one handoff-ready Markdown artifact under
spdd/discovery/and recommend exactly one next phase fromspdd-plan,spdd-story, orspdd-analysis. references/local-commands/command contracts: SpecArk-authored command text now lives outside slimSKILL.mdwrappers for local workflow skills such asspdd-orchestrator,spdd-discovery, andspdd-plan.
Changed
spdd-orchestratorandspdd-plannow use thin wrappers that delegate behavior to local command files, keeping callable skill files smaller while preserving the same workflow contracts.- Bundle validation now checks discovery skill files, docs visibility, Discovery Brief naming support, and local command wrapper conventions.
- Public docs now place discovery before planning and story creation, while noting that discovery is manually invoked until separate orchestrator discovery support lands.
Notes
spdd-discoverydoes not generate stories, analysis, prompts, code, or tests. It preserves early product context and hands that context to the next SPDD phase.- Automatic orchestrator routing into discovery remains a separate follow-up.
0.3.1 — Documentation sync skill
Added
spdd-doc-syncskill: standalone maintenance skill that keeps human-facing in-repo documentation aligned after any workflow, skill, or behavior change. Accepts a free-form change description and uses an explicit routing table to determine which documentation surfaces to update — and which to leave untouched.- Affected-vs-untouched report: every invocation produces a structured report listing which documentation files were updated with reasons, and which were deliberately skipped with reasons. Supports auditable, traceable doc changes.
docs/skills/spdd-doc-sync.md: dedicated documentation page explaining when and how to use the skill and how it differs fromspdd-sync.- Maintenance skills section added to
docs/skills/index.md.
Notes
spdd-doc-syncis a terminal maintenance step — it always emitsnext_phase: completeand does not chain intospdd-generate.spdd-doc-syncandspdd-syncare complementary: runspdd-syncto keep the structured prompt accurate, thenspdd-doc-syncto propagate those changes into the human-facing docs.- The skill enforces a hard boundary: it never reads or writes files under
spdd/prompt/(that path is owned byspdd-sync).
0.3.0 — Session health & UX polish
Added
spdd-session-healthskill: pre-flight context-health check that assesses whether the current conversation is in good shape to execute the next SPDD phase.- Orchestrator pre-flight integration:
spdd-orchestratorautomatically invokesspdd-session-healthwhen multiple input artifacts are detected or a prior phase was completed in the current session. - Cowork compatibility surfaced in docs — no plugin changes required.
- SpecArk logo in the docs navbar, favicon, and README.
Changed
- Home page, installation pages, and palette refreshed and unified for clarity.
- Platform logos resized to small clickable thumbnails on the home and installation pages.
Fixed
- Docs base path corrected after the repository rename so GitHub Pages assets resolve correctly.
- Stale submodule install instructions corrected in README and release notes.
Notes
spdd-planskill (added during the 0.2.0 cycle but not documented at the time) is now in the workflow table — seeplugins/specark/CLAUDE.md.
0.2.0 — Claude Code support
Added
- Claude Code platform support: SpecArk now works with both Codex and Claude Code via the standard plugin marketplace mechanism.
plugins/specark/.claude-plugin/plugin.json: Claude Code plugin manifest, mirroring the existing.codex-plugin/plugin.jsonlayout..claude-plugin/marketplace.json: repo-root marketplace file forclaude plugin marketplace add.docs/installation-claude-code.md: step-by-step Claude Code install guide (claude plugin marketplace add maxgda/spec-ark+claude plugin install specark).- Real Anthropic and OpenAI inline SVG logo marks on the docs home page and installation pages.
- GitHub quick-link action button on the docs home page.
- VitePress code-group tabs on Getting Started and First Feature Tutorial for side-by-side Codex / Claude Code syntax.
Changed
- All 9
SKILL.mdfiles: addeddisable-model-invocation: trueto frontmatter and removed platform-specific "Codex skill" wording from descriptions. docs/index.md: platform compatibility badges with real logo marks; GitHub action button; Codex-only framing removed.docs/installation.md: restructured into Codex and Claude Code sections with inline SVG platform icons.docs/getting-started.md,docs/first-feature.md:/specark:skill-nameinvocation syntax shown for Claude Code.docs/.vitepress/config.mjs: added Installation — Claude Code to the Overview sidebar.plugins/specark/CLAUDE.md: simplified to a reference overview; install entry-point role moved to.claude-plugin/plugin.json.README.md: Claude Code startup section updated toclaude plugin marketplace add/claude plugin install.validate_plugin_bundle.py: checks both.claude-plugin/plugin.jsonand.claude-plugin/marketplace.json, plusdisable-model-invocation: truein every skill.
Notes
- The Codex install path (
codex plugin marketplace add) is unchanged. agents/openai.yamlfiles are unchanged — Claude Code does not consume them.
0.1.x — Initial documentation refresh
The documentation refresh in this cycle focused on:
- separating user-facing usage guidance from maintainer-only notes
- making the
READMEa cleaner front door for installation, workflow execution, token savings, and test generation - improving VitePress navigation so release preparation material is easier to find
Release checklist
Before publishing a new version:
- Run
python3 plugins/specark/scripts/validate_plugin_bundle.py. - Run
npm run docs:build. - Review
CHANGELOG.mdand add the final versioned entry. - Review
docs/release-notes.mdand move any unreleased release summary into the final versioned section. - Verify
plugins/specark/.codex-plugin/plugin.jsonandplugins/specark/.claude-plugin/plugin.jsonhave correct version and metadata. - Restart Codex and verify the plugin is discoverable from
.agents/plugins/marketplace.json. - Test Claude Code installation:
claude plugin marketplace add maxgda/spec-ark && claude plugin install specark. - Tag the release after validation is clean.
Suggested release summary format
Use a short structure for each release:
Added
- new skills, scripts, references, or docs sections
Changed
- workflow behavior, prompts, docs structure, or packaging metadata
Fixed
- validator issues, broken docs links, packaging mistakes, or routing problems
Notes
- anything maintainers or adopters should watch during upgrade
Relationship to changelog
Use CHANGELOG.md as the versioned source of truth.
Use this page for:
- release preparation guidance
- summary expectations
- consistent notes structure across releases