@hegemonart/get-design-done 1.59.4 → 1.59.6

package/.claude-plugin/marketplace.json

@@ -4,15 +4,15 @@
     "name": "hegemonart"
   },
   "metadata": {
-    "description": "Get Design Done — 5-stage agent-orchestrated design pipeline (Brief → Explore → Plan → Design → Verify) for AI coding agents. 64 agents, 95 skills, 39 connection integrations, two MCP servers, opt-in SQLite state backbone, bidirectional Figma write-back, and a reflector-driven self-improvement loop. Cross-runtime install for Claude Code, Codex, Cursor, OpenCode, Gemini, and more.",
-    "version": "1.59.4"
+    "description": "Get Design Done — 5-stage agent-orchestrated design pipeline (Brief → Explore → Plan → Design → Verify) for AI coding agents. 64 agents, 96 skills, 39 connection integrations, two MCP servers, opt-in SQLite state backbone, bidirectional Figma write-back, and a reflector-driven self-improvement loop. Cross-runtime install for Claude Code, Codex, Cursor, OpenCode, Gemini, and more.",
+    "version": "1.59.6"
   },
   "plugins": [
     {
       "name": "get-design-done",
       "source": "./",
-      "description": "Agent-orchestrated 5-stage design pipeline (Brief → Explore → Plan → Design → Verify) for AI coding agents. 64 specialized agents, 95 skills, 39 connection integrations (Figma, Refero, Preview, Storybook, Chromatic, Graphify, Linear, Jira, Notion, …), bidirectional Figma write-back, queryable intel store, opt-in SQLite state backbone, and a reflector-driven self-improvement loop. Two MCP servers (gdd-state for typed STATE mutators, gdd-mcp for 13 read-only project-priming tools), tier-aware routing with cost telemetry, and defense-in-depth hooks (protected paths, MCP circuit breaker, injection scanner, budget enforcer). Cross-runtime install for Claude Code, Codex, Cursor, OpenCode, Gemini, Copilot, and more.",
-      "version": "1.59.4",
+      "description": "Agent-orchestrated 5-stage design pipeline (Brief → Explore → Plan → Design → Verify) for AI coding agents. 64 specialized agents, 96 skills, 39 connection integrations (Figma, Refero, Preview, Storybook, Chromatic, Graphify, Linear, Jira, Notion, …), bidirectional Figma write-back, queryable intel store, opt-in SQLite state backbone, and a reflector-driven self-improvement loop. Two MCP servers (gdd-state for typed STATE mutators, gdd-mcp for 13 read-only project-priming tools), tier-aware routing with cost telemetry, and defense-in-depth hooks (protected paths, MCP circuit breaker, injection scanner, budget enforcer). Cross-runtime install for Claude Code, Codex, Cursor, OpenCode, Gemini, Copilot, and more.",
+      "version": "1.59.6",
       "author": {
         "name": "hegemonart"
       },

package/.claude-plugin/plugin.json

@@ -1,8 +1,8 @@
 {
   "name": "get-design-done",
   "short_name": "gdd",
-  "version": "1.59.4",
-  "description": "Agent-orchestrated 5-stage design pipeline (Brief → Explore → Plan → Design → Verify) for AI coding agents. 64 specialized agents, 95 skills, 39 connection integrations (Figma, Refero, Preview, Storybook, Chromatic, Graphify, Linear, Jira, Notion, …), bidirectional Figma write-back, queryable intel store for O(1) design-surface lookups, opt-in SQLite state backbone, and a reflector-driven self-improvement loop. Two MCP servers (`gdd-state` for typed STATE mutators, `gdd-mcp` for 13 read-only project-priming tools), tier-aware agent routing with cost telemetry, defense-in-depth hooks (protected paths, MCP circuit breaker, injection scanner, budget enforcer), and a cross-runtime install layer for Claude Code, Codex, Cursor, OpenCode, Gemini, Copilot, and more.",
+  "version": "1.59.6",
+  "description": "Agent-orchestrated 5-stage design pipeline (Brief → Explore → Plan → Design → Verify) for AI coding agents. 64 specialized agents, 96 skills, 39 connection integrations (Figma, Refero, Preview, Storybook, Chromatic, Graphify, Linear, Jira, Notion, …), bidirectional Figma write-back, queryable intel store for O(1) design-surface lookups, opt-in SQLite state backbone, and a reflector-driven self-improvement loop. Two MCP servers (`gdd-state` for typed STATE mutators, `gdd-mcp` for 13 read-only project-priming tools), tier-aware agent routing with cost telemetry, defense-in-depth hooks (protected paths, MCP circuit breaker, injection scanner, budget enforcer), and a cross-runtime install layer for Claude Code, Codex, Cursor, OpenCode, Gemini, Copilot, and more.",
   "author": {
     "name": "hegemonart",
     "url": "https://github.com/hegemonart"

package/CHANGELOG.md

@@ -4,6 +4,68 @@ All notable changes to get-design-done are documented here. Versions follow [sem
 
 ---
 
+## [1.59.6] - 2026-06-05
+
+Sixth and final point release of the **v1.59 "Audit Closeout & Honesty Pass"** milestone. Product-surface docs, the handoff skill, and the fact-gate. **Milestone complete.**
+
+### Added
+
+- **Real `/gdd:handoff` skill** (`skills/handoff/`), replacing the phantom command the README and Jump Mode advertised with no skill behind it. A thin wrapper that initializes a cycle from a Claude Design bundle and routes to verify. (Skill count 95 to 96.)
+- **`docs/KNOWLEDGE-BASE.md`**, a navigable index of the reference library, linked from the README.
+
+### Changed
+
+- **Product-first README** (623 to 402 lines): rewritten from the readme-redesign draft, leading with what the tool does and how to start. The 6-badge block and the i18n language nav are preserved verbatim; release-chronicle prose now lives only in this CHANGELOG.
+- **Honest multi-runtime claim**: the full 14-runtime breadth list carries an explicit caveat (the full design-quality gate is Claude Code; hooks are Claude-only; MCP works on ~3 runtimes; the rest are install-supported but untested).
+- **i18n lag banners** added to all 6 localized READMEs (translations may lag the canonical English).
+
+### Fixed (fact-gate)
+
+- Removed 3 phantom commands from the README that had no backing skill (`/gdd:scan`, `/gdd:set-profile`, `/gdd:run-skill`); every `/gdd:` command in the README now resolves to a real skill in the manifest.
+- Corrected stale counts to the shipped reality: **64 agents, 96 skills, 39 connection integrations, 13 read-only MCP tools, 14 runtimes**, and the knowledge-base "222 reference files" figure to the actual 189.
+
+### Milestone v1.59 complete
+
+Six point releases (1.59.1 to 1.59.6) closed out the v1.58.1 audit: MCP servers auto-register, the connection / skill / agent counts are honest and gated, agent model/tier routing is coherent, the skill-authoring surface ships to npm, the Batch-H polish landed, and every command / count / runtime claim in the README now resolves against the code.
+
+### Breaking changes
+
+None.
+
+5,070/5,070 tests pass.
+
+---
+
+## [1.59.5] - 2026-06-05
+
+Fifth point release of the **v1.59 "Audit Closeout & Honesty Pass"** milestone. Batch H polish + runtime-model provenance.
+
+### Added
+
+- **Risk-calibration now learns from the bandit feedback loop.** The bandit's post-spawn `recordOutcome` also updates the per-agent calibration table (best-effort), so calibration tracks the same signal that drives routing. (H2)
+- **Runtime-model provenance guard.** `budget-enforcer` no longer applies a HARD budget cap from a BYOK / unverified runtime-model row (it degrades to advisory), and the schema records the provenance. This closes the risk of an unverified placeholder tier hard-blocking a user. (P1)
+
+### Fixed
+
+- **Cursor installs no longer drop co-located skill reference files.** `installMultiArtifact` now carries a skill's sibling `*-procedure.md` reference files alongside `SKILL.md` for Cursor's flat layout, with symmetric uninstall cleanup. (H6)
+
+### Hardening
+
+- Pinned the state backup-guard rotation cap (10 slots) + non-empty corruption check with a dedicated test, and verified `state-store.migrate()` is async with complete JSDoc. (H5/H7)
+
+### Notes
+
+- Batch-H item H8 (the `composes_with` composition-graph backfill) is owned by Phase 58, not this release; no double-implementation here.
+- Documented follow-ups (each a larger-than-a-slice scope cut): calibration `detectDrift` to reflector consumption, SQLite-header corruption detection in the backup guard, the cursor sibling-carry generalized to all flat-layout runtimes, and the structural `status:` key on runtime-model entries (currently blocked by the parser's allowed-key enforcement).
+
+### Breaking changes
+
+None.
+
+5,070/5,070 tests pass.
+
+---
+
 ## [1.59.4] - 2026-06-04
 
 Fourth point release of the **v1.59 "Audit Closeout & Honesty Pass"** milestone. Skill-surface + build hygiene.