v3.0.0: Competitive constructed, sideboard strategy, Moxfield search

[j4th]

Summary

• 4 metagame workflow tools — metagame_snapshot, archetype_decklist, archetype_comparison, format_entry_guide (MTGGoldfish + Spicerack backends)
• 3 sideboard workflow tools — suggest_sideboard, sideboard_guide, sideboard_matrix (bulk data + MTGGoldfish)
• 2 Moxfield provider tools — moxfield_search_decks, moxfield_user_decks
• Fuzzy matching utility for archetype and matchup name resolution
• Comprehensive documentation alignment — README, COOKBOOK, ARCHITECTURE, CONTRIBUTING, TOOL_DESIGN all updated to 69 tools / 19 prompts / 21 resources
• Changelog + release notes with friend-friendly TL;DR

By the Numbers

Metric	v2.2.0	v3.0.0
Tools	60	69
Prompts	17	19
Resources	18	21
Tests	1160	1340
Coverage	88%	88%
Data sources	5	8

Test plan

• mise run check — 1340 tests passed, 88% coverage
• mise run test:live — 35/35 live smoke tests passed
• Review changelog entry completeness
• Review release notes TL;DR tone
• Verify no stale version strings remain

🤖 Generated with Claude Code

[claude]

Claude finished @j4th's task in 4m 43s —— View job

---

PR Review — v3.0.0 Release

• Gather context (PR diff, changed files)
• Pass 1: Correctness & Convention Compliance
• Pass 2: Silent Failure Hunting
• Pass 3: Test Coverage Gaps
• Pass 4: Documentation & Infrastructure Completeness

---

PR Review Summary

Verdict: APPROVE

This is a pure release engineering PR — version bumps, changelog, release notes, and documentation alignment. No Python source files changed. The code for the new features (metagame workflows, sideboard tools, Moxfield search) was already on main from prior PRs; this PR packages them as a formal v3.0.0 milestone.

Critical Issues (must fix)

None.

Important Issues (should fix)

• docs/RELEASE_NOTES_v3.md and PR body — data source count discrepancy (confidence: 82)

  The two v2.2.0 baseline counts disagree:

  • PR body table: Data sources | 5 | 8
  • docs/RELEASE_NOTES_v3.md table: Data sources | 6 | 8

  Neither is consistent with the codebase: by v2.2.0 (60 tools), Moxfield (decklist + deck_info), Spicerack (3 tools), and MTGGoldfish (4 tools) were all already mounted — so v2.2.0 had 8 data sources, same as v3.0.0. This release adds 9 new tools but zero new backends. The table comparison vs. v2.2.0 is misleading; it appears to be comparing against an earlier pre-MTGGoldfish baseline.

  Since the PR is already merged this is informational, but worth correcting in RELEASE_NOTES_v3.md to avoid confusion for users checking the changelog. Fix this →

Completeness

• Version strings: ✅ pyproject.toml and both server.json version fields updated to 3.0.0
• Env vars in server.json: ✅ MTG_MCP_ENABLE_MOXFIELD, MTG_MCP_ENABLE_SPICERACK, MTG_MCP_ENABLE_MTGGOLDFISH all added
• Tool/prompt/resource counts: ✅ 69 tools / 19 prompts / 21 resources consistent across README.md, CONTRIBUTING.md, COOKBOOK.md, ARCHITECTURE.md, TOOL_DESIGN.md
• Security fix: ✅ pygments>=2.20.0 in pyproject.toml, uv.lock pinned to 2.20.0 (CVE-2026-4539)
• Docs: ✅ CHANGELOG.md, RELEASE_NOTES_v3.md, docs/README.md (index link), COOKBOOK.md (Constructed recipes), all updated
• Smoke tests: ✅ .claude/skills/smoke-test/SKILL.md includes metagame_snapshot, archetype_decklist, suggest_sideboard, sideboard_guide, sideboard_matrix, moxfield_search_decks
• PEP 8 comment fixes (Task 7): ✅ Correctly skipped — all three flagged lines (building.py:34, constructed.py:69, cache.py:24) are continuation lines within multi-line comments, not standalone sentences
• Tests: ✅ test_orchestrator_e2e.py:28 asserts exactly 69 tools; PR reports 1340 tests at 88% coverage, 35/35 live tests pass

What's Good

• Meticulous count alignment across every doc file — no stale "51 tools" / "17 prompts" / "18 resources" references found
• Well-structured CHANGELOG entry with proper categorization (Added/Changed/Fixed/Security)
• Release notes written at the right abstraction level — feature-oriented TL;DR, not a diff dump
• Planning docs committed to docs/superpowers/ for reproducibility (future releases can reference the pattern)
• Pygments CVE addressed with both a constraint update and a lockfile pin