v1.9.0

What's Changed

• docs: Context7 config, llms.txt and agent-discovery notes by @humanity4ai in #5
• feat(mcp): hosted endpoint parity (per-tool schemas, outputSchema, prompts) + npm in server.json; v1.9.0 by @humanity4ai in #6

Full Changelog: v1.8.2...v1.9.0

v1.8.2 — SEO follow-up

Serves og.png from the repo root (Vercel static root), fixes the last stale '16 tools' badge, and updates constitution/quick-reference from the retired SSE endpoint to the streamable-HTTP /api/mcp endpoint.

v1.8.1 — SEO & discoverability

Open Graph/Twitter/JSON-LD on the landing page and the hosted MCP docs page, a real PNG og:image, registry title/description, robots.txt + sitemap.xml, and GitHub repo metadata. Fixes the stale '16 tools' meta description.

v1.8.0 — vdd_inspect (16 → 15 MCP tools)

Folds vdd_trace + vdd_analyze into one read-only vdd_inspect tool (scope enum: project|feature), reducing the MCP surface to 15 tools and lifting Glama's Tool Count Appropriateness dimension from 3 to 5. Engine/CLI keep vdd trace / vdd analyze.

v1.7.1 — graceful degradation for missing selector arguments

Selector-based tools no longer hard-fail when an argument is omitted: they auto-select the only spec directory (feature) or return an actionable success payload (output.needsInput). Fixes the Agent Status reliability index dropping to Degraded. Applied in the engine and the hosted Vercel handler, with regression tests.

v1.7.0 — SSE transport retired

Removed

• The legacy SSE endpoint is retired. api/sse.js → non-routed api/_vdd-rpc.js; /api/sse now returns an HTTP 308 redirect to /api/mcp. server.json advertises only the streamable-http remote. This eliminates the recurring Task timed out after 30 seconds errors on /api/sse.

Added

• Hosted-surface regression tests (16 tools, titles, annotation classification, and PHASE_META parity between the Vercel handler and the engine).
• CI build-test job: pnpm -r build → pnpm check → pnpm test → node --check on the handlers.
• Browser docs page at GET /api/mcp.

Fixed

• Removed the ignored memory setting from vercel.json (build warning gone).
• openWorldHint: false on local-only tools; only strategize and clone are open-world.
• Annotation accuracy carried from ebc4d5e (clarify/implement/amend read-only; validate additive).

Docs

• README, SKILL, AGENTS, index.html now point at /api/mcp only.

v1.6.0 — Streamable HTTP transport + Glama registry

What's new

• Streamable HTTP transport (packages/vdd-mcp/src/http.ts + http-entry.ts): the same 17 VDD tools now serve over the MCP Streamable HTTP transport (port 3000), alongside the existing stdio entrypoint — enables one-click hosting/deploy on the Glama registry.
• Glama registry metadata (glama.json) and a purpose-built glama/Dockerfile for reproducible sandbox builds + introspection.
• Tool titles + MCP annotation hints (readOnlyHint/destructiveHint/idempotentHint/openWorldHint) on all 17 tools.

Fixed

• validate discovers feature dirs under vdd/specs/ and no longer overwrites a hand-authored impact-report.md.

MCP

17 tools (was 16 — vdd_e2e miscounted), stdio + Streamable HTTP, no API key required.

v1.5.3 — Full Consistency, Zero Benchmark Gaps

[1.5.3] — 2026-08-10

Fixed

• README.md: Anti-pattern (23→24), domain primers (4→6), added scripts/ and .github/ to structure
• AGENTS.md: Expanded directory map with all 7 primers, added constitution.md and scripts/
• references/INDEX.md: Gate checks (110→113), anti-patterns (16→24), added missing primers
• references/quick-reference.md: Added human-factors + verification-toolchain to Domain Primers table
• vdd/docs/tutorial.md, vdd/strategy.md, vdd/tactics.md: Version and count fixes

Added

• domain-primers/safety-critical.md: FMEA/FTA primer, safety integrity levels, VDD integration
• references/compliance-evidence.md: Evidence maps for DO-178C, IEC 62304, CMMI, ISO 29148

Verified

• Full-repo consistency audit: 113 gates, 24 anti-patterns, 7 domain primers all consistent
• Best-practice benchmark: 47/47 (100%), 11 exceeded, 0 gaps

v1.5.0 — Full-Spectrum, Quality-First, Best-Practice Aligned

Vision Driven Design v1.5.0

57 files · 8,141 lines · 37 action items · 11 strategic pillars · 6 domain primers

20 commits since v1.0.0. VDD has evolved from an initial methodology into a comprehensive,
best-practice-aligned, full-spectrum development system.

What's New in v1.5.0

11 Strategic Pillars
Adoption, Authority, Community, Proof, Accessibility, GitHub Growth, Agent Integration,
Spectrum Coverage, Substance Enforcement, Benchmark Excellence, Quality-First Toolchain

6 Domain Primers
WebApp · Data Storage (SurrealDB 3.x, PostgreSQL, Neo4j, Redis, MongoDB) · ETL · Infrastructure ·
Human Factors (behavioral economics, cognitive load, habit formation) ·
Verification Toolchain (Playwright, Browserless, Sentry, CI/CD)

4 Scripts

• scripts/install.sh — One-line curl installer
• scripts/vdd-agent-setup.sh — Multi-agent config generator (OpenCode, Claude Code, Cursor, Copilot)
• scripts/vdd-detect.sh — Project convention auto-detector
• scripts/vdd-substance-audit.sh — Low-substance artifact detection

CI/CD

• .github/workflows/vdd-quality-gates.yml — Impact chain verification, detection, progress tracking
• vdd/docs/ci-quality-pipeline.md — 9-stage pipeline template (lint → typecheck → unit → e2e → visual → a11y → security → perf → deploy → sentry)

Documentation (8 Guides)

• Tutorial — 30-minute walkthrough
• Comparison — VDD vs SDD vs vibe coding vs TDD
• Vision Canvas — 5-min template for non-technical visionaries
• Best-Practice Benchmark — 7 standards, 42/42 matched
• Star-Growth Campaign
• Community Engagement
• Telemetry Plan
• Agent SDK Concept

Dogfood Repo
github.com/simonplmak-cloud/vdd-dogfood-task-tracker — Complete VDD chain + working Next.js app built with VDD

Benchmark Validation
42/42 industry criteria matched across NASA SE, CMMI REQM, DO-178C, IEC 62304, DORA 2025, ISO 29148, GitHub Spec Kit

Install

curl -sSL https://raw.githubusercontent.com/simonplmak-cloud/vision-driven-design/main/scripts/install.sh | bash

v1.0.0 — Initial Release

Vision Driven Design v1.0.0

Initial release of the VDD methodology and OpenCode skill.

What's Included

• 8-phase chain: Vision → Strategy → Tactics → Specs → Plan → Tasks → Implement → Validate
• 7 bi-directional gates with 113 verification checks
• 4 domain primers: WebApp, Data Storage, ETL, Infrastructure
• 11 artifact templates with impact chain headers
• Full-auto mode: human provides vision, AI drives the rest
• Recursive Goldratt S&T decomposition at every phase junction
• 23 documented anti-patterns with symptoms and fixes

Installation

git clone https://github.com/simonplmak-cloud/vision-driven-design.git ~/.config/opencode/skills/vision-driven-design/

Documentation

See README.md and references/.