@hegemonart/get-design-done 1.0.7 → 1.13.3

package/.claude-plugin/marketplace.json

@@ -5,14 +5,14 @@
   },
   "metadata": {
     "description": "Get Design Done — 5-stage agent-orchestrated design pipeline with 9 connections, handoff-first workflow, bidirectional Figma write-back, 22+ specialized agents, queryable knowledge layer (intel store, dependency analysis, learnings extraction), and a self-improvement loop (reflector, frontmatter + budget feedback, global-skills layer). Ships with a full CI/CD pipeline (Node 22/24 × Linux/macOS/Windows) and release automation (auto-tag + GitHub Release + release-time smoke test).",
-    "version": "1.0.7"
+    "version": "1.13.3"
   },
   "plugins": [
     {
       "name": "get-design-done",
       "source": "./",
       "description": "Agent-orchestrated 5-stage design pipeline: Brief → Explore → Plan → Design → Verify. 22+ specialized agents, 9 connections (Figma, Refero, Preview, Storybook, Chromatic, Figma Writer, Graphify, Pinterest, Claude Design), Claude Design handoff, bidirectional Figma write-back, and a queryable intel store (.design/intel/) for dependency and learnings queries. Standalone commands: style, darkmode, compare, figma-write, graphify, handoff, analyze-dependencies, skill-manifest, extract-learnings. Embeds NNG heuristics, WCAG thresholds, typographic systems, motion framework, and anti-pattern catalog. Ships with a full CI/CD pipeline (Node 22/24 × Linux/macOS/Windows) and release automation. Optimization layer (v1.0.4.1, retroactive): gdd-router + gdd-cache-manager skills, PreToolUse budget-enforcer hook, tier-aware agent frontmatter, lazy checker gates, streaming synthesizer, /gdd:warm-cache + /gdd:optimize commands, and cost telemetry at .design/telemetry/costs.jsonl — targeting 50-70% per-task token-cost reduction with no quality-floor regression.",
-      "version": "1.0.7",
+      "version": "1.13.3",
       "author": {
         "name": "hegemonart"
       },

package/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "get-design-done",
   "short_name": "gdd",
-  "version": "1.0.7",
+  "version": "1.13.3",
   "description": "Agent-orchestrated 5-stage design pipeline: Brief → Explore → Plan → Design → Verify. 22+ specialized agents, 9 connections (Figma, Refero, Preview, Storybook, Chromatic, Figma Writer, Graphify, Pinterest, Claude Design), handoff-first workflow via Claude Design bundles, bidirectional Figma write-back (annotations, Code Connect), queryable intel store (`.design/intel/`) for O(1) design surface lookups, and self-improvement loop (reflector agent, frontmatter + budget feedback, global-skills layer at `~/.claude/gdd/global-skills/`). Standalone commands: style, darkmode, compare, figma-write, graphify, handoff, analyze-dependencies, skill-manifest, extract-learnings, reflect, apply-reflections. Embeds NNG heuristics, WCAG thresholds, typographic systems, motion framework, and anti-pattern catalog. Ships with a full CI/CD pipeline (Node 22/24 × Linux/macOS/Windows, lint + schema + frontmatter + stale-ref + shellcheck + gitleaks + injection-scan + blocking size-budget) and release automation (auto-tag + GitHub Release + release-time smoke test). Optimization layer (v1.0.4.1, retroactive): gdd-router + gdd-cache-manager skills, PreToolUse budget-enforcer hook, tier-aware agent frontmatter, lazy checker gates, streaming synthesizer, /gdd:warm-cache + /gdd:optimize commands, and cost telemetry at .design/telemetry/costs.jsonl — targeting 50-70% per-task token-cost reduction with no quality-floor regression.",
   "author": {
     "name": "hegemonart",

package/CHANGELOG.md

@@ -4,6 +4,88 @@ All notable changes to get-design-done are documented here. Versions follow [sem
 
 ---
 
+## [1.13.3] — 2026-04-19
+
+### Added — Phase 13.3: Plugin Update Checker
+
+- `hooks/update-check.sh` — SessionStart Bash hook; 24h-cached unauthenticated `GET /repos/hegemonart/get-design-done/releases/latest`; classifies semver delta (major / minor / patch / off-cadence); respects `.design/STATE.md` stage guard (suppresses nudge during `plan`, `design`, `verify`); respects per-version dismissal from `.design/config.json`; silent-on-failure by policy (exit 0 on every error path); BASH_SOURCE guard makes the script source-safe for self-test.
+- `hooks/hooks.json` — registers `update-check.sh` as a second SessionStart command alongside `bootstrap.sh` (run order: bootstrap → update-check).
+- `agents/design-update-checker.md` — Haiku-tier enrichment agent (Phase 10.1 cost governance). Invoked only by `/gdd:check-update --prompt` to produce a 3–5-line "what this release changes for you" summary from the cached release body. Reads-only, inline-text-only output, ends with `## UPDATE-CHECKER COMPLETE`.
+- `skills/check-update/SKILL.md` — `/gdd:check-update` manual slash command. Flags: `--refresh` (bypass 24h TTL, re-fetch now), `--dismiss` (write `update_dismissed: "<tag>"` atomically to `.design/config.json` via env-prefix python3 heredoc, preserves all pre-existing keys), `--prompt` (spawn the enrichment agent). Default (no flag) prints cached state.
+- `.design/update-cache.json` — per-user-project runtime cache (written by the hook). Shape: `{checked_at, current_tag, latest_tag, delta, is_newer, changelog_excerpt}`; 500-char excerpt from release body.
+- `.design/update-available.md` — per-user-project runtime rendered banner. Written only when all four gates pass (cache is_newer=true AND stage ∉ {plan,design,verify} AND not dismissed AND latest_tag parsed successfully). Consumed by safe-window skills via `[ -f .design/update-available.md ] && cat .design/update-available.md`.
+- `reference/schemas/config.schema.json` — adds optional `update_dismissed: string` property.
+- Safe-window surfaces: `/gdd:progress`, `/gdd:health`, `/gdd:help`, post-closeout of `/gdd:ship`, `/gdd:complete-cycle`, `/gdd:audit` — each appends the one-line banner-cat tail before its completion marker. Mid-pipeline skills (`brief`, `explore`, `plan`, `design`, `verify`) explicitly NOT modified.
+- `skills/audit/SKILL.md` — `tools:` list extended with `Bash` to enable the banner-cat tail (previously: `Read, Write, Task, Glob`).
+- Root `SKILL.md` — `check-update` registered in argument-hint alternation.
+- `test-fixture/baselines/phase-13.3/` — regression baseline lock for v1.13.3.
+- `test-fixture/baselines/current/agent-list.txt` — appended `design-update-checker.md` in sorted position.
+- `test-fixture/baselines/current/skill-list.txt` — appended `check-update` in sorted position.
+
+### Changed
+
+- Plugin version: 1.0.7.2 → 1.13.3 (per the new milestone.phase.sub-phase versioning scheme — MAJOR=milestone, MINOR=phase, PATCH=sub-phase). Phase 13.3 lands as a valid semver; release workflow auto-tags and `npm publish` succeeds.
+
+### Design principles (Phase 13.3)
+
+- **Never auto-updates.** The checker only surfaces a nudge; `/gdd:update` remains the explicit user action.
+- **Never interrupts critical work.** State-machine guard suppresses the nudge during mid-pipeline stages (`plan`, `design`, `verify`); banner renders only in the 6 documented safe windows.
+- **Silent-on-failure.** Network timeout, malformed JSON, missing plugin.json, unwritable `.design/` — every path exits 0 without printing to stderr during normal SessionStart.
+- **No telemetry.** Unauthenticated GitHub Releases fetch; no phone-home, no tracking, no tokens in code.
+
+---
+
+## [1.0.7.2] — 2026-04-19
+
+### Added — Phase 13.2: External Authority Watcher
+
+- `reference/authority-feeds.md` — curated whitelist of 26 design-authority feeds grouped by kind (5 spec sources, 8 component systems, 3 research institutions, 10 named practitioners, user-extensible Are.na channels) with an explicit `## Rejected kinds` assertion block covering Dribbble, Behance, LinkedIn, generic Medium topic feeds, and trending aggregators.
+- `reference/schemas/authority-snapshot.schema.json` — Draft-07 JSON Schema validating `.design/authority-snapshot.json` shape (version const 1, feeds keyed by id, 64-hex sha256 hash pattern, `maxItems: 200` per-feed entry cap enforcing D-14 retention at validation time).
+- `agents/design-authority-watcher.md` — Sonnet-tier, parallel-safe watcher agent. Fetches feeds via WebFetch, diffs against snapshot, classifies new entries into five buckets (`spec-change`, `heuristic-update`, `pattern-guidance`, `craft-tip`, `skip`) with a one-sentence rationale per entry. First run seeds the snapshot silently; `--since <date>` is the escape hatch for surfacing a backlog.
+- `skills/watch-authorities/SKILL.md` — `/gdd:watch-authorities` command with `--refresh`, `--since <date>`, `--feed <name>`, `--schedule <weekly|daily|monthly>` flags. Mutual-exclusion rules between `--refresh` and `--since`; `--schedule` registers a cron via the `scheduled-tasks` MCP when connected (weekly=`0 9 * * 1`, daily=`0 9 * * *`, monthly=`0 9 1 * *`) with graceful exit-0 fallback on MCP absence.
+- `scripts/tests/test-authority-rejected-kinds.sh` — CI test enforcing the anti-slop thesis structurally. Splits `reference/authority-feeds.md` at the `## Rejected kinds` heading and greps the active section for `dribbble.com` / `behance.net` / `linkedin.com` / `medium.com/topic` / trending-aggregator hostnames; exits non-zero on any match.
+- `scripts/tests/test-authority-watcher-diff.sh` — structural-only v1 of the watcher-diff test. Asserts fixture presence, baseline existence, D-21 classification-heading vocabulary, and exact count consistency between the header's "N entries surfaced" figure and the number of bulleted entries across classification sections. Full end-to-end byte diff against a live watcher run is deferred until the Claude Code agent runtime is available in CI.
+- `test-fixture/authority-feeds/` — four frozen mock feeds (WAI-ARIA APG Atom, Radix Primitives release Atom, NN/g articles RSS, Are.na channel JSON) with deterministic timestamps for byte-stable CI diff testing. Exercises all four non-practitioner source kinds against every branch of the D-17 classification decision table.
+- `test-fixture/baselines/phase-13.2/authority-report.expected.md` — frozen regression baseline for the watcher-diff test (9 entries across 4 feeds: spec-change=2, heuristic-update=1, pattern-guidance=4, craft-tip=2, skip=0).
+
+### Changed
+
+- `skills/reflect/SKILL.md` step 3 "Build required-reading list" — appended `.design/authority-report.md`. Single-line addition; `agents/design-reflector.md` itself is byte-identical since phase start (D-25 reflector-non-modification invariant preserved).
+- Root `SKILL.md` — `watch-authorities` registered in argument-hint alternation, the maintenance Command Reference table, and the Jump Mode routing block alongside `/gdd:reflect` and `/gdd:apply-reflections`.
+- `scripts/validate-schemas.cjs` — wired `authority-snapshot` into the `PAIRS` array and invoked `ajv-cli` with `-c ajv-formats` so `format: "date-time"` declarations are enforced rather than rejected under ajv strict mode. No-op for existing schemas (none declared formats).
+- `tests/agent-size-budget.test.cjs` — added `M: 300` tier to `TIER_LIMITS` for Worker-tier agents (between `S: 150` and `LARGE: 350`); accommodates CONTEXT D-05's "body ≈ 200–300 lines" target with modest headroom.
+- `test-fixture/baselines/phase-6/agent-list.txt` — appended `design-authority-watcher.md` in sorted position.
+- `test-fixture/baselines/phase-6/skill-list.txt` — appended `watch-authorities` in sorted position.
+- Plugin version: 1.0.7 → 1.0.7.2 (off-cadence decimal patch; 1.0.7.1 was a CHANGELOG-only label consumed by Phase 13.1 Figma MCP consolidation and did not bump manifests). Does not shift the Phase 14 → v1.0.8 cadence.
+
+---
+
+## [1.0.7.1] — 2026-04-19
+
+### Changed — Phase 13.1: Figma MCP Consolidation
+- Collapsed the dual Figma MCP setup (local `figma-desktop` for reads + remote `figma` for writes) into the single remote `figma` MCP, which exposes full read parity (`get_metadata`, `get_design_context`, `get_variable_defs`, `get_screenshot`) alongside `use_figma` for writes.
+- Rewrote `connections/figma.md` to cover both reads and writes; deleted `connections/figma-writer.md` (folded into the unified spec).
+- Migrated every `mcp__figma-desktop__*` tool reference to `mcp__figma__*` across skills (`scan`, `discover`, `explore`, `design`), agents (`design-figma-writer`, `design-context-builder`, `design-discussant`, `token-mapper`), and `connections/connections.md` capability matrix + probe block.
+- Collapsed STATE.md `<connections>` schema from `figma: … / figma_writer: …` to a single `figma:` key. The remote MCP is one server — one probe, one status.
+- Updated capability matrix in `connections/connections.md`: a single `Figma` row now declares write-back under the `design` column (FWR-01..04).
+- Regenerated `test-fixture/baselines/phase-6/connection-list.txt` to drop the deleted `figma-writer.md` entry.
+
+### Migration
+- Install the remote Figma MCP (one command; replaces both prior installs):
+  ```
+  claude mcp add figma --transport http https://mcp.figma.com/v1/sse
+  ```
+- Optionally remove the old desktop MCP after upgrading:
+  ```
+  claude mcp remove figma-desktop
+  ```
+- No command or flag renames. The `design-figma-writer` agent keeps its name and proposal→confirm UX unchanged.
+
+### Off-cadence note
+- Labelled **v1.0.7.1** per the decimal-phase convention (established by Phase 10.1 → v1.0.4.1). The label is CHANGELOG-only: `package.json`, `.claude-plugin/plugin.json`, and `.claude-plugin/marketplace.json` remain at `1.0.7` so the release workflow does not auto-tag. The next integer bump (Phase 14 → v1.0.8) will absorb these changes implicitly. This matches how Phase 10.1 was handled.
+
+---
+
 ## [1.0.7] — 2026-04-18
 
 ### Added — Phase 13: CI/CD

package/README.md

@@ -68,7 +68,9 @@ Get Design Done fixes that. It's the context engineering layer for design work i
 
 ## Who This Is For
 
-Designers and design-engineers who want Claude Code to do serious design work — and expect it to respect the design system, the brief, and WCAG.
+Anyone shipping UI with Claude Code who expects the output to actually hold up — engineers, designers, design-engineers, solo founders. If you care that tokens match, contrast passes WCAG, and the result ties back to what you asked for, this is for you.
+
+You don't need to be a designer. The pipeline carries the design expertise so you don't have to — it extracts the system, grounds in references, verifies against the brief, and catches the things people usually miss.
 
 Built-in quality gates catch real problems: Handoff Faithfulness scoring on Claude Design bundles, contrast audits across the full palette × surface matrix, anti-pattern detection from the NNG catalog, dark-mode architecture verification, and motion-system consistency checks.
 
@@ -285,14 +287,13 @@ Every stage uses the same pattern: a thin orchestrator skill spawns specialized
 
 All agents carry `default-tier: haiku|sonnet|opus` + `tier-rationale` in frontmatter. Every agent opens with `@reference/shared-preamble.md` so the first agent in a session pays full cost and the rest ride Anthropic's 5-minute prompt cache.
 
-### 9 Tool Connections
+### 8 Tool Connections
 
 Every connection is optional. The pipeline degrades gracefully — a grep-based fallback exists for every missing tool.
 
 | Connection | Type | Purpose |
 |-----------|------|---------|
-| Figma Desktop | MCP (`mcp__figma-desktop__*`) | Token extraction, design context pre-population |
-| Figma Writer | MCP (`mcp__figma__use_figma`) | Write decisions back to Figma (annotate, tokenize, Code Connect) |
+| Figma | MCP (`mcp__figma__*`, remote) | Token extraction, design context pre-population, write-back via `use_figma` (annotate, tokenize, Code Connect) |
 | Refero | MCP (`mcp__refero__*`) | Reference design search during exploration |
 | Pinterest | MCP (`mcp__mcp-pinterest__*`) | Visual inspiration boards alongside Refero |
 | Preview (Playwright) | MCP (`mcp__Claude_Preview__*`) | Live page screenshots for visual verification |
@@ -331,6 +332,47 @@ After each design cycle, `/gdd:reflect` reads `.design/learnings/`, `.design/tel
 
 ---
 
+## Authority Watcher
+
+Subscribe to a curated whitelist of design-authority sources, diff it against a snapshot, and feed only genuinely new, classified entries into the Self-Improvement reflector. Authority monitoring — not trend watching.
+
+```bash
+# On-demand diff + classify
+/gdd:watch-authorities
+
+# Force re-seed the snapshot (recovery for a corrupted snapshot)
+/gdd:watch-authorities --refresh
+
+# Surface backlog since a specific date
+/gdd:watch-authorities --since 2026-01-01
+
+# Limit to a single feed (debugging)
+/gdd:watch-authorities --feed wai-aria-apg
+
+# Schedule recurring runs (requires the scheduled-tasks MCP)
+/gdd:watch-authorities --schedule weekly
+```
+
+### What the whitelist covers
+
+See [`reference/authority-feeds.md`](reference/authority-feeds.md). 26 curated feeds grouped by kind:
+
+- **Spec sources** — WAI-ARIA APG, Material 3, Apple HIG, Fluent 2, W3C Design Tokens CG
+- **Component systems** — Radix, shadcn/ui, Polaris, Carbon, Primer, Atlassian, Ant, Mantine
+- **Research institutions** — Nielsen Norman Group, Laws of UX, Baymard
+- **Named practitioners** — 10 writers filtered for spec-adjacent, durable, original analysis
+- **User-added Are.na channels** — extensibility point; add your own via a PR to the Are.na section, no config file or schema editing required
+
+### What is explicitly rejected
+
+No Dribbble. No Behance. No LinkedIn. No generic "trending" aggregators. See `reference/authority-feeds.md` §"Rejected kinds" — enforced structurally by `scripts/tests/test-authority-rejected-kinds.sh`, which greps the active whitelist for rejected hostnames and fails CI on any match. The anti-slop thesis is a test, not a comment.
+
+### How the report feeds reflection
+
+The watcher writes `.design/authority-report.md` — new entries classified into five buckets (`spec-change`, `heuristic-update`, `pattern-guidance`, `craft-tip`, `skip`) with a one-sentence rationale each. `/gdd:reflect` reads the report alongside internal telemetry and proposes reference-file updates. Nothing auto-ships — you review every proposal via `/gdd:apply-reflections`.
+
+---
+
 ## Commands
 
 All commands use the `/gdd:` namespace.
@@ -443,13 +485,15 @@ Full command reference with argument specs: [`SKILL.md`](SKILL.md).
 
 All connections are optional — the pipeline degrades gracefully when any connection is unavailable.
 
-### Figma MCP
+### Figma MCP (reads + writes)
 
-When the official Figma Desktop MCP is active, `explore` reads Figma variables and pre-populates design decisions from your file. Falls back to code-only analysis. Setup: [`connections/figma.md`](connections/figma.md).
+One remote MCP covers both reads and writes as of v1.0.7.1. When it is active, `explore` reads Figma variables and pre-populates design decisions from your file, and `design-figma-writer` writes decisions back via `use_figma` — annotates frames, tokenizes local styles, registers Code Connect mappings. Proposal → confirm discipline with `--dry-run` and `--confirm-shared` guards. Falls back to code-only analysis when the MCP is not configured. One install command unlocks both:
 
-### Figma Writer MCP
+```
+claude mcp add figma --transport http https://mcp.figma.com/v1/sse
+```
 
-`design-figma-writer` writes decisions back to Figma via the remote `use_figma` MCP — annotates frames, tokenizes local styles, registers Code Connect mappings. Proposal → confirm discipline with `--dry-run` and `--confirm-shared` guards. Setup: [`connections/figma-writer.md`](connections/figma-writer.md).
+Setup: [`connections/figma.md`](connections/figma.md). If you previously installed the local `figma-desktop` MCP, you can remove it — its read tools are now exposed on the remote server.
 
 ### Refero MCP
 
@@ -590,7 +634,7 @@ Every `/gdd:*` command and agent spawn passes through a cross-cutting optimizati
 - **Streaming synthesizer** (`skills/synthesize/`) — N parallel-mapper outputs collapse through a single Haiku call before returning to main context.
 - **Cost telemetry** — `.design/telemetry/costs.jsonl` appends one row per spawn decision: `{ts, agent, tier, tokens_in, tokens_out, cache_hit, est_cost_usd, cycle, phase}`. Aggregated to `.design/agent-metrics.json`. Consumed by `/gdd:optimize` and `design-reflector`.
 
-Regression baseline: [`test-fixture/baselines/phase-10.1/`](test-fixture/baselines/phase-10.1/) — methodology README + `pre-baseline-cost-report.md` + `cost-report.md`. Phase 13 CI diffs against `cost-report.md` to catch regressions.
+CI diffs against a locked cost baseline on every push to catch regressions.
 
 ---
 
@@ -623,19 +667,18 @@ lint → validate → test (matrix) → security + size-budget
 - **Agent hygiene** — frontmatter completeness, line-count tier budgets, required-reading path validity, `/gdd:` namespace consistency
 - **System contracts** — config schema, command↔skill parity, hooks integrity, atomic writes to `.design/STATE.md`, frontmatter parser edge cases, model-profile resolution, `/gdd:health` output shape, worktree safety, semver bump sequence, STATE-TEMPLATE drift
 - **Pipeline + data** — end-to-end smoke on `test-fixture/`, mapper JSON-schema validation (tokens, components, a11y, motion, hierarchy), parallelism-engine decision table, `Touches:` field parsing, cycle lifecycle, `.design/intel/` incremental-update correctness, regression-baseline drift detector
-- **Feature correctness** — `/gdd:sketch` variant determinism, 9-connection probe contracts with mocked MCPs, `design-figma-writer` dry-run discipline, `design-reflector` proposal-only shape, deprecated-name redirects, NNG heuristic coverage, injection-scanner hook behavior, optimization-layer schema enforcement
+- **Feature correctness** — `/gdd:sketch` variant determinism, 8-connection probe contracts with mocked MCPs, `design-figma-writer` dry-run discipline, `design-reflector` proposal-only shape, deprecated-name redirects, NNG heuristic coverage, injection-scanner hook behavior, optimization-layer schema enforcement
 
 ### Release automation
 
-`.github/workflows/release.yml` auto-tags and publishes a GitHub Release when `.claude-plugin/plugin.json` version changes. Release body is extracted from the matching `CHANGELOG.md` section. A release-time smoke test (`scripts/release-smoke-test.cjs`) diffs a fresh checkout against `test-fixture/baselines/phase-13/` before the release publishes.
+`.github/workflows/release.yml` auto-tags and publishes a GitHub Release when `.claude-plugin/plugin.json` version changes. Release body is extracted from the matching `CHANGELOG.md` section. A release-time smoke test diffs a fresh checkout against a locked baseline before the release publishes.
 
-From **v1.0.6 forward**, every PR MUST pass `npm test` before merging to `main`. Baselines in `test-fixture/baselines/phase-<N>/` lock each phase's structural state — re-lock explicitly rather than relaxing tests. See [`CONTRIBUTING.md`](CONTRIBUTING.md).
+Every PR must pass `npm test` before merging to `main`. See [`CONTRIBUTING.md`](CONTRIBUTING.md) for the re-lock procedure when baselines change.
 
 ---
 
-## Distribution
+## What ships with the plugin
 
-**Ships with the plugin:**
 - `.claude-plugin/plugin.json`, `.claude-plugin/marketplace.json` — manifest
 - `SKILL.md` — root pipeline router
 - `skills/` — 55 stage + standalone skills
@@ -644,11 +687,6 @@ From **v1.0.6 forward**, every PR MUST pass `npm test` before merging to `main`.
 - `reference/` — curated design reference (shared preamble, model tiers, model prices, schemas, DEPRECATIONS, config schema)
 - `hooks/`, `scripts/bootstrap.sh`
 
-**Dev-only (gitignored, not distributed):**
-- `.planning/` — GSD planning artifacts for GDD's own development
-- `.claude/memory/` — session-level memory
-- `.claude/settings.local.json`
-
 ---
 
 ## Develop locally

package/SKILL.md

@@ -2,7 +2,7 @@
 name: get-design-done
 short_name: gdd
 description: "Master design pipeline for Claude Code. 5-stage workflow: Brief → Explore → Plan → Design → Verify. Run 'brief' first in any new project to capture the design problem, then 'explore' to inventory the codebase and interview for context. Invoke without arguments for status and auto-routing."
-argument-hint: "[brief|explore|plan|design|verify|handoff|map|next|help|status|style|darkmode|compare|figma-write|graphify|discuss|list-assumptions|progress|health|todo|stats|note|plant-seed|add-backlog|review-backlog|scan|discover|settings|update|reapply-patches|audit|pause|resume|new-cycle|debug|quick|new-project|complete-cycle|fast|do|ship|undo|pr-branch|sketch|sketch-wrap-up|spike|spike-wrap-up|reflect|apply-reflections|analyze-dependencies|extract-learnings|skill-manifest|warm-cache|optimize|cache-manager]"
+argument-hint: "[brief|explore|plan|design|verify|handoff|map|next|help|status|style|darkmode|compare|figma-write|graphify|discuss|list-assumptions|progress|health|todo|stats|note|plant-seed|add-backlog|review-backlog|scan|discover|settings|update|reapply-patches|audit|pause|resume|new-cycle|debug|quick|new-project|complete-cycle|fast|do|ship|undo|pr-branch|sketch|sketch-wrap-up|spike|spike-wrap-up|reflect|apply-reflections|analyze-dependencies|extract-learnings|skill-manifest|warm-cache|optimize|cache-manager|watch-authorities|check-update]"
 user-invocable: true
 ---
 
@@ -83,6 +83,7 @@ Each stage produces artifacts in `.design/` inside the current project.
 | `analyze-dependencies [--slice <name>]` | `get-design-done:analyze-dependencies` | Query the `.design/intel/` store — dependency slices, graph queries, phase-scoped reads |
 | `extract-learnings [--cycle <slug>]` | `get-design-done:extract-learnings` | Extract decisions, lessons, patterns, and surprises from a completed cycle → `.design/cycles/<slug>/LEARNINGS.md` |
 | `skill-manifest [--refresh]` | `get-design-done:skill-manifest` | List or refresh the local skill manifest used by the router for discovery |
+| `watch-authorities [--refresh] [--since <date>] [--feed <name>] [--schedule <cadence>]` | `get-design-done:gdd-watch-authorities` | Run design-authority-watcher — fetch curated feeds, diff snapshot, classify new entries → `.design/authority-report.md` (consumed by `/gdd:reflect`) |
 
 ## Handoff Routing
 
@@ -125,9 +126,9 @@ Each stage produces artifacts in `.design/` inside the current project.
 
 6. **Optional: Bidirectional write-back** (post-verify, offered to user)
    After verify completes without FAIL-level gaps:
-   - Check STATE.md `<connections>` for `figma_writer`
-   - `figma_writer: not_configured` → skip (no offer)
-   - `figma_writer: available` → offer: "Write implementation status back to Figma? (annotates frames + Code Connect mappings)"
+   - Check STATE.md `<connections>` for `figma:` (the unified remote MCP covers both reads and writes as of v1.0.7.1)
+   - `figma: not_configured` or `figma: unavailable` → skip (no offer)
+   - `figma: available` → offer: "Write implementation status back to Figma? (annotates frames + Code Connect mappings)"
      Options: [yes, write back] [dry-run, show proposal only] [skip]
    - If yes or dry-run: spawn `agents/design-figma-writer.md` with `mode: implementation-status`, `dry_run: <true|false>`
 
@@ -200,6 +201,7 @@ If `$ARGUMENTS` is a stage or command name — invoke it directly, no state chec
 /gdd:audit              → Skill("get-design-done:gdd-audit")
 /gdd:reflect            → Skill("get-design-done:gdd-reflect")
 /gdd:apply-reflections  → Skill("get-design-done:gdd-apply-reflections")
+/gdd:watch-authorities  → Skill("get-design-done:gdd-watch-authorities")
 /gdd:pause              → Skill("get-design-done:gdd-pause")
 /gdd:resume          → Skill("get-design-done:gdd-resume")
 # --- Lifecycle ---

package/agents/design-authority-watcher.md

@@ -0,0 +1,208 @@
+---
+name: design-authority-watcher
+description: Fetches a curated whitelist of design-authority feeds, diffs against .design/authority-snapshot.json, classifies new entries into five buckets, emits .design/authority-report.md. Spawned by /gdd:watch-authorities.
+tools: Read, Write, WebFetch, Bash, Grep, Glob
+color: blue
+model: inherit
+default-tier: sonnet
+tier-rationale: "Network fetch + per-entry classification is open-ended pattern recognition; Haiku misclassifies spec-vs-opinion boundary, Opus overkill for a weekly run."
+size_budget: M
+parallel-safe: always
+typical-duration-seconds: 90
+reads-only: false
+writes:
+  - ".design/authority-snapshot.json"
+  - ".design/authority-report.md"
+---
+
+@reference/shared-preamble.md
+
+# design-authority-watcher
+
+## Role
+
+You are the network-fetching agent for the authority-watcher phase. You read the whitelist at `reference/authority-feeds.md`, fetch each feed via `WebFetch` (or the Are.na v2 API for `kind: arena` entries), diff against `.design/authority-snapshot.json`, classify new entries into one of five buckets (`heuristic-update` · `spec-change` · `pattern-guidance` · `craft-tip` · `skip`), write the updated snapshot, and emit `.design/authority-report.md`. You never modify `agents/design-reflector.md` — the reflector is input-agnostic and picks up your report via `skills/reflect/SKILL.md` step 3 (wired in Plan 13.2-03). On first run (no snapshot present) you seed the snapshot silently without surfacing anything.
+
+## Required Reading
+
+The orchestrating skill supplies a `<required_reading>` block in the prompt. Read every listed file before acting — this is mandatory. Minimum expected inputs (skip gracefully if absent, note what is missing):
+
+- `reference/authority-feeds.md` — the curated whitelist you fetch from.
+- `.design/authority-snapshot.json` — prior snapshot (absent = first run per D-15).
+- `.design/STATE.md` — for cycle slug if present (non-fatal if absent).
+
+## Flags
+
+Flags are supplied by the orchestrating skill in the prompt (the skill parses `/gdd:watch-authorities` user arguments):
+
+- `--refresh` → re-seed snapshot from current feed state without surfacing anything (D-23 recovery mode; behaves identically to first run).
+- `--since <ISO8601 date>` → surface entries whose `published` date is newer than the given boundary regardless of snapshot state (D-15 escape hatch + D-23 backlog surfacing).
+- `--feed <feed-id>` → fetch only the single named feed (debugging / spot-check).
+
+The `--schedule` flag is handled by the skill (Plan 13.2-03), not by this agent. If you receive it, ignore.
+
+## Step 1 — Load Whitelist
+
+Read `reference/authority-feeds.md`. Parse each feed entry (lines matching `^- \*\*\[.+\]\(https?://`) into a tuple `{ title, homepage, kind, url, cadence-hint, rationale, feed-id }`. Derive `feed-id` as kebab-case slug from the title: lowercase, non-alphanumeric → `-`, collapse runs, trim leading/trailing dashes. Entries inside HTML comments (`<!-- ... -->`) are placeholders — ignore. Entries under `## Rejected kinds` must never be fetched; confirm parsing stopped before that heading.
+
+If `--feed <feed-id>` is set, filter the list to the single matching entry. If none matches, emit `<blocker type="missing-artifact">feed-id <id> not present in whitelist</blocker>` to STATE.md and terminate with `## WATCH COMPLETE`.
+
+## Step 2 — Load Prior Snapshot
+
+Read `.design/authority-snapshot.json`.
+
+- If absent: set `first_run = true`, initialize in-memory `feeds = {}`.
+- If present: parse JSON. Validate `version === 1`. On mismatch, emit `<blocker type="contract-violation">authority-snapshot.json version mismatch: expected 1</blocker>` to STATE.md and do NOT continue to Step 6 (write).
+
+If `--refresh` is set, behave as if `first_run = true` regardless of prior snapshot state.
+
+## Step 3 — Fetch Loop
+
+For each feed in the filtered list, fetch content. Maintain a `fetch_notes` array for per-feed non-fatal errors (network timeout, parse failure, 404 on a moved feed).
+
+**`kind: arena`** — GET `https://api.are.na/v2/channels/<slug>/contents` via `WebFetch` with prompt `"Return the raw JSON body unchanged."`. Parse JSON. For each content block, build an entry:
+
+```
+id       = String(block.id)
+title    = block.title || block.generated_title || "Untitled"
+summary  = (block.description || "").slice(0, 2000)
+permalink = block.class === "Link" ? block.source.url : "https://are.na/block/" + block.id
+published = block.created_at   // used only for --since filtering
+```
+
+**All other kinds** — GET the feed URL via `WebFetch` with prompt:
+
+> Return the feed as a structured list of entries with fields: id (use guid or link), title, summary (use description/summary/content:encoded, strip HTML tags), link, published (ISO8601 if available). Prefer Atom fields over RSS when both appear.
+
+Parse the structured reply into entries with the same field names as the arena branch.
+
+**Polite-crawl:** between requests to the **same host** (by `URL.host`), sleep 250ms (D-11). Distinct hosts may fetch back-to-back without delay. A per-feed inline `min-delay-ms:` override in the whitelist (if present) supersedes the default.
+
+**Errors are non-fatal.** On WebFetch or parse failure, push `{ feed-id, error: "<one-sentence>" }` into `fetch_notes` and continue. A single failing feed must not block the other ~25.
+
+## Step 4 — Diff
+
+For each feed's newly-fetched entries, compute a content hash:
+
+```
+hash = sha256(title + "\n" + summary)
+```
+
+Use `Bash` to invoke `printf '%s\n%s' "$title" "$summary" | shasum -a 256 | awk '{print $1}'` (or the Node `crypto.createHash('sha256').update(title+"\n"+summary).digest('hex')` equivalent). Output MUST be a 64-char lowercase hex string — the schema at `reference/schemas/authority-snapshot.schema.json` enforces `^[0-9a-f]{64}$`.
+
+**New-entry rule** (D-13):
+- Entry is new if its `id` is not present in `prior.feeds[feed-id].entries`, OR
+- Entry is new if its `id` IS present but the `hash` differs from the stored one (content changed).
+
+**`--since <date>` modifier:** also mark entries whose `published > since` as new, independent of snapshot membership. This is the backlog escape hatch.
+
+**First-run / refresh short-circuit:** if `first_run === true` (either initial run or `--refresh`) AND `--since` is absent, classify nothing — accumulate every fetched entry directly into the new snapshot and skip Step 5. Proceed to Step 6.
+
+## Step 5 — Classify
+
+Apply the decision table below to each new entry. Emit `{ ...entry, classification, rationale }` where `rationale` is a ≤1-sentence deterministic trace of which rule matched (e.g., "title matched `/added|updated|removed/i` → spec-change"). Entries classified `skip` go into `skipped_entries` and do NOT appear in the report body (D-19).
+
+**Classification decision table (D-17):**
+
+| Source kind | Default classification |
+|---|---|
+| `spec-source` | `spec-change` if title matches `/(added|updated|deprecated|removed|new)/i` else `pattern-guidance` |
+| `component-system` | `pattern-guidance` if title matches new-component regex (`/(new|add(ed)?|introduc(e|ing))/i` AND contains a component noun like button/dialog/menu/modal/card/tooltip/popover/select/combobox/etc.) else `craft-tip` |
+| `research` | `heuristic-update` if title or summary mentions `/principle|law|heuristic|usability finding/i` else `craft-tip` |
+| `named-practitioner` | `craft-tip` by default; upgrade to `pattern-guidance` if the entry's link points to a spec-source host (`w3.org`, `developer.apple.com`, `m3.material.io`, `fluent2.microsoft.design`) |
+| `arena` | `pattern-guidance` (user-curated references are pattern material by construction) |
+| any, flagged promo/newsletter/ad or matching skip-regex `/(sponsor(ed)?\|newsletter\|promo(tion)?\|\[ad\]\|subscribe|unsubscribe|webinar)/i` in title | `skip` (takes precedence over all above) |
+
+The skip row is evaluated LAST and overrides the kind-based row — a component-system release titled "Sponsored: shipping our new sponsor tier" still ends up `skip`.
+
+## Step 6 — Write Snapshot
+
+For each feed, merge the newly-fetched entries into `feeds[feed-id].entries`:
+- Preserve the prior entries for ids not seen this run (stale entries persist until pruned).
+- For ids seen this run, overwrite the prior record with `{ id, hash }` from the fresh fetch.
+- Append order: existing retained entries first (oldest → newest), then new arrivals.
+- **Prune: keep only the last 200 entries per feed** (D-14). This is a hard cap; the schema at `reference/schemas/authority-snapshot.schema.json` rejects >200 via `maxItems:200`, so pruning MUST happen before the write call.
+
+Set `feeds[feed-id].last_fetched_at` to the current ISO8601 UTC timestamp. Set top-level `generated_at` to the same. Serialize with 2-space indentation.
+
+**Pre-write contract check:** before calling `Write`, walk the serialized object and verify:
+1. `version === 1`.
+2. Every `feeds[*].entries[*].hash` matches `^[0-9a-f]{64}$`.
+3. No feed's `entries.length > 200`.
+
+If any check fails, emit `<blocker type="contract-violation">` to STATE.md with the offending path and do NOT write. Terminate with `## WATCH COMPLETE`.
+
+On pass, write `.design/authority-snapshot.json`.
+
+## Step 7 — Write Report
+
+Write `.design/authority-report.md`. Overwritten every run.
+
+**First-run / refresh mode** (either `first_run` true or `--refresh` set, and `--since` absent):
+
+```markdown
+# Authority Report — <ISO date>
+
+Seeded snapshot for N feeds — next run will surface new entries.
+```
+
+No sections, no footer. Terminate the file after the single sentence.
+
+**Normal mode** — header, sections, footer:
+
+```markdown
+# Authority Report — <ISO date>
+
+N entries surfaced across M feeds. K skipped.
+
+## spec-change (X)
+- **[Title](url)** — feed: <feed-title> — *<rationale sentence>*
+
+## heuristic-update (X)
+- **[Title](url)** — feed: <feed-title> — *<rationale sentence>*
+
+## pattern-guidance (X)
+- **[Title](url)** — feed: <feed-title> — *<rationale sentence>*
+
+## craft-tip (X)
+- **[Title](url)** — feed: <feed-title> — *<rationale sentence>*
+
+---
+
+**Skipped:** K entries (see `.design/authority-snapshot.json` for the full trail).
+```
+
+**Rules:**
+- Classification sections ordered by weight: `spec-change` → `heuristic-update` → `pattern-guidance` → `craft-tip` (D-21).
+- Omit a section entirely when its count is zero (signal density).
+- The **Skipped** footer line is ALWAYS present — even when K=0 — for Plan 13.2-04 diff-test determinism.
+- If `fetch_notes` is non-empty, append a `Fetch notes:` block after the Skipped line, one bullet per note:
+  ```markdown
+
+  Fetch notes:
+  - <feed-id>: <one-sentence error>
+  ```
+- Entry line format is exact: `- **[Title](url)** — feed: <feed-title> — *<rationale>*`. Em-dash (`—`), italicized rationale, no trailing period unless the rationale itself ends one.
+
+## Step 8 — Output
+
+Emit a single-line summary to stdout:
+
+- **Normal mode:** `Surfaced N entries across M feeds. K skipped. See .design/authority-report.md.`
+- **First-run / refresh mode:** `Seeded snapshot for N feeds — next run will surface new entries.`
+
+## Do Not
+
+- Do NOT modify `agents/design-reflector.md`. Reflector integration is Plan 13.2-03's scope and lives in `skills/reflect/SKILL.md` only.
+- Do NOT fetch URLs that are not listed in `reference/authority-feeds.md`. The whitelist is the allow-list.
+- Do NOT spawn subagents — you have no `Task` tool for a reason.
+- Do NOT commit on behalf of the user. `.design/authority-snapshot.json` and `.design/authority-report.md` both live under gitignored `.design/`.
+- Do NOT write outside your declared `writes:` list. If work appears to require another write, stop and return a `<blocker>`.
+
+## Completion
+
+On contract violation (schema mismatch, hash format violation, over-200 entries) emit a `<blocker>` to STATE.md per the preamble protocol. Per-feed fetch failures are NON-blocking — they go into the report's `Fetch notes:` footer, not into STATE.md.
+
+Terminate every response with:
+
+## WATCH COMPLETE

package/agents/design-context-builder.md

@@ -1,7 +1,7 @@
 ---
 name: design-context-builder
 description: Detects existing design system state via grep/glob, runs discovery interview asking ONLY unanswered questions, produces .design/DESIGN-CONTEXT.md. Spawned by the discover stage.
-tools: Read, Write, Bash, Grep, Glob, mcp__figma-desktop__get_variable_defs, mcp__figma-desktop__get_metadata, mcp__refero__search
+tools: Read, Write, Bash, Grep, Glob, mcp__figma__get_variable_defs, mcp__figma__get_metadata, mcp__refero__search
 required_reading:
   - connections/storybook.md
 color: blue
@@ -50,10 +50,10 @@ The orchestrating stage supplies a `<required_reading>` block in the prompt. Rea
 **ToolSearch first.** Figma tools may be in the deferred tool set — calling them without a prior ToolSearch fails silently.
 
 ```
-ToolSearch({ query: "figma-desktop", max_results: 10 })
+ToolSearch({ query: "select:mcp__figma__get_variable_defs", max_results: 1 })
 ```
 
-Then call `mcp__figma-desktop__get_variable_defs` (no arguments — returns all variables in the active Figma file).
+Then call `mcp__figma__get_variable_defs` (no arguments — returns all variables in the active Figma file).
 
 > If `get_variable_defs` errors (most commonly because no Figma file is open): skip Step 0 entirely AND update `.design/STATE.md` `<connections>` to `figma: unavailable`. Proceed to Step 1 with no pre-populated decisions.
 

package/agents/design-discussant.md

@@ -30,7 +30,7 @@ The spawning prompt supplies `<required_reading>`. Read every listed file before
 
 ## Step 0 — Context pre-load (Figma only, optional)
 
-If `<connections>` in STATE.md shows `figma: available`, ToolSearch `figma-desktop` and call `mcp__figma-desktop__get_variable_defs`. For each returned variable, draft a *tentative* D-XX decision (mark "tentative — confirm with user"). Silently skip on any error. Do NOT grep the codebase.
+If `<connections>` in STATE.md shows `figma: available`, `ToolSearch({ query: "select:mcp__figma__get_variable_defs", max_results: 1 })` and call `mcp__figma__get_variable_defs`. For each returned variable, draft a *tentative* D-XX decision (mark "tentative — confirm with user"). Silently skip on any error. Do NOT grep the codebase.
 
 ## Step 1 — Mode dispatch
 

package/agents/design-figma-writer.md

@@ -1,7 +1,7 @@
 ---
 name: design-figma-writer
 description: Writes design decisions back to Figma — annotations, token bindings, Code Connect mappings, and implementation-status write-back. Operates in proposal→confirm mode by default. Accepts --dry-run (emit proposal without executing) and --confirm-shared (required for writes to team library components).
-tools: Read, Write, Bash, Grep, Glob, mcp__figma__use_figma, mcp__figma-desktop__get_variable_defs, mcp__figma-desktop__get_metadata
+tools: Read, Write, Bash, Grep, Glob, mcp__figma__use_figma, mcp__figma__get_variable_defs, mcp__figma__get_metadata
 color: purple
 model: inherit
 default-tier: sonnet
@@ -34,7 +34,7 @@ ToolSearch({ query: "select:mcp__figma__use_figma", max_results: 1 })
 → Non-empty     → proceed to Step 1
 ```
 
-Note: `mcp__figma__use_figma` is the remote Figma MCP (registered as server "figma"). Distinct from `mcp__figma-desktop__*` (desktop MCP, read-only in this pipeline). If only desktop MCP is present and remote is absent, STOP with the note above.
+Note: `mcp__figma__use_figma` is the remote Figma MCP (registered as server `figma`). Reads (`mcp__figma__get_metadata`, `mcp__figma__get_variable_defs`) live on the same server. As of v1.0.7.1, there is no separate read-only desktop MCP — the remote MCP is the single supported Figma connection.
 
 ---
 
@@ -72,15 +72,15 @@ Read `.design/DESIGN-CONTEXT.md`. Extract the relevant data for the selected mod
 - For `tokenize`: color/spacing/type literal values that could map to Figma variables — look for hex values, spacing scales, and typography sizes in the decisions section
 - For `mappings`: component names and their source file paths — look for component listings, file paths, and implementation references
 
-Also read the active Figma file structure using the desktop MCP (reads are always desktop, writes are always remote):
+Also read the active Figma file structure using the remote MCP (reads and writes share the same server):
 
 ```
-ToolSearch({ query: "figma-desktop", max_results: 10 })
-mcp__figma-desktop__get_metadata()    // lightweight layer outline
-mcp__figma-desktop__get_variable_defs()   // for tokenize mode — variable names and values
+ToolSearch({ query: "select:mcp__figma__get_metadata,mcp__figma__get_variable_defs", max_results: 2 })
+mcp__figma__get_metadata()       // lightweight layer outline
+mcp__figma__get_variable_defs()  // for tokenize mode — variable names and values
 ```
 
-If `get_metadata` errors (no file open), write: "No Figma file is open. Open the target file in the Figma desktop app and retry." and STOP.
+If `get_metadata` errors (no file accessible), write: "No Figma file is accessible. Open the target file in Figma and retry." and STOP.
 
 ---
 
@@ -157,7 +157,7 @@ Wait for user response. If response is not "yes", STOP with "Cancelled."
 
 ## Step 5 — Execute Writes
 
-For each operation in the proposal, call `mcp__figma__use_figma` with the appropriate operation payload. Remote MCP only — never use desktop MCP for writes.
+For each operation in the proposal, call `mcp__figma__use_figma` with the appropriate operation payload.
 
 For `annotate`: