Batch 11 — Skill Marketplaces & Curated Plugin Collections (Tier B)
Roster (10)
| slug | stars | distribution | cli_binary | local_ui | orchestration | multi_model | tier |
|---|---|---|---|---|---|---|---|
| claude-skills-mcp | 387 | mcp-server | yes (claude-skills-mcp) | none | none | no | A |
| claude-skills-marketplace-mhatting | 586 | claude-plugin | no | none | sequential | no | A |
| context-engineering-kit-neolabhq | 1042 | claude-plugin | no | none | hierarchical | no | A |
| muratcankoylan-skills | 16030 | claude-plugin | no | none | none | no | A |
| sgaunet-plugins | 13 | claude-plugin | no | none | sequential | yes (partial) | A |
| anilcancakir-cc-plugins | 6 | claude-plugin | no | none | sequential | no | A |
| glebis-skills | 225 | claude-plugin | yes (gpt_image_2.py) | none | hierarchical | no | A |
| behisecc-claude-skills | 9246 | methodology-doc | no | none | none | no | B (awesome-list) |
| droids-curtis | 253 | standalone-repo | no | none | sequential | yes (gpt-5) | A |
| harness-engineering-skills-phlegon | 17 | skill-pack | yes (bun harness:*) | none | hierarchical | unknown | A |
Catalog Size Table (Organizing Principle Per Repo)
| slug | skills | agents | hooks | MCP | commands | organizing principle |
|---|---|---|---|---|---|---|
| claude-skills-mcp | 0 | 0 | 0 | 2 (3 tools) | 0 | Semantic search over external skill catalogs via MCP proxy architecture |
| claude-skills-marketplace-mhatting | 18 | 1 | 0 | 1 (optional) | 0 | Plugin-per-domain with optional code execution MCP for token efficiency |
| context-engineering-kit-neolabhq | 3 (reflexion) | 5+ (sdd/sadd) | 2 | 0 | 1 | Granular per-plugin install with TypeScript/Bun Stop-hook automated adversarial reflection |
| muratcankoylan-skills | 15 | 0 | 0 | 0 | 0 | Educational stack: foundational→architectural→operational→harness theory |
| sgaunet-plugins | 2 | 14 | 0 | 1 (optional) | 8 | Domain-specialist agents with per-agent tool restrictions and model assignments |
| anilcancakir-cc-plugins | 4 | 3 | 7 | 0 | unknown | Phase-aware context compaction (5 lifecycle hooks per session) |
| glebis-skills | 48 | 3 (tdd) | 0 | 0 | 0 | Personal productivity OS with flagship multi-agent TDD (context isolation between phases) |
| behisecc-claude-skills | 0 | 0 | 0 | 0 | 0 | Awesome-list index of 50+ external skills across 14 categories |
| droids-curtis | 0 | 29 | 0 | 0 | 0 | Forced-context production droids for Cloudflare Workers/Neon/tRPC stack |
| harness-engineering-skills-phlegon | 2 | 19 | 0 | 0 | 8 (bun) | PRD-to-code delivery loop with 3-tier ceremony, filesystem-enforced phase gates |
Intra-batch Patterns
All 10 repos are Claude Code plugin-compatible, but the organizing philosophy varies dramatically: some are discovery layers (claude-skills-mcp, behisecc-claude-skills), some are knowledge libraries (muratcankoylan-skills), and some are production delivery systems (harness-engineering-skills-phlegon, glebis-skills TDD). The batch shows a clear split between breadth-first personal skill libraries (glebis at 48 plugins, behisecc as an index) and depth-first specialist collections (sgaunet's 14 domain experts, droids-curtis's 29 forced-context production agents). Hook usage is concentrated in two repos (anilcancakir-cc-plugins with 7 hooks, context-engineering-kit-neolabhq with 2 TypeScript Bun hooks), while the remaining 8 repos use zero lifecycle hooks. The muratcankoylan-skills collection stands alone as an academic-quality educational library cited in peer-reviewed papers — unique in the entire analyzed corpus. All repos except behisecc-claude-skills write some form of SKILL.md or persona-md files, confirming the dominance of the Anthropic skill format across the personal-curation category.
Most Interesting Finds
glebis-skills (TDD skill): The multi-agent TDD orchestration with architecturally enforced context isolation between Test Writer and Implementer subagents is the most principled application of information-boundary engineering in any personal skill repo. The
--dry-runmode that renders all subagent prompts with character counts before execution is novel — it treats the orchestration pipeline itself as testable software.droids-curtis: Explicit
model: gpt-5routing for the Cloudflare deployment specialist droid is the only cross-provider (Anthropic → OpenAI) model assignment in the entire analyzed corpus, suggesting the author pragmatically benchmarked both providers per task domain.anilcancakir-cc-plugins (strategic-compact): The
PreCompactlifecycle hook that fires before every compaction event and injects context-preservation guidance is architecturally unique — no seed and no other batch item treats the compaction transition as a first-class workflow event.harness-engineering-skills-phlegon: Filesystem-enforced phase gates (
bun harness:advanceonly advances when required artifacts exist on disk) is the only code-enforced (not prompt-enforced) phase state machine in the corpus. Combined with a dedicatedcontext-compactor.mdagent, this is the most production-engineering-complete skill pack in the batch.
Items Written as Tier C Stubs
None. All 10 repos had sufficient public material for full 11-file analysis.
Note: behisecc-claude-skills is functionally Tier B content (awesome-list with no original engineering) but was analyzed fully because all 11 files were writable from the README alone.
Cross-References Discovered
- behisecc-claude-skills explicitly indexes
obra/superpowersskills (test-driven-development, using-git-worktrees, finishing-a-development-branch, brainstorming) — direct cross-reference to seed framework - behisecc-claude-skills lists
K-Dense-AI/claude-scientific-skillsas a collection — the same source indexed by claude-skills-mcp (batch item 1) - glebis-skills TDD design credits
tdd-guard(nizos/tdd-guard, hook-enforced TDD) and Matt Pocock's TDD skill - context-engineering-kit-neolabhq is listed in "Mentioned in Awesome Claude Code" and is cross-referenced from behisecc-claude-skills 14-category list
- muratcankoylan-skills is cited in the same academic survey paper that established harness engineering as a field — conceptually linked to harness-engineering-skills-phlegon (which teaches harness patterns vs implements them)
- claude-skills-mcp is the MCP delivery layer for K-Dense-AI/claude-scientific-skills — a producer/consumer relationship between two repos in this batch's ecosystem