🆕 July 18, 2026 — 4 New Hermes-Agent Skill Repos¶
Date: July 18, 2026 New Repos: 4 | New Skills: 17+ | Combined Stars: 26
Evening discovery sweep across GitHub repos tagged hermes-agent and agent-skills created July 17-18, 2026. Cross-referenced against 324 existing catalog entries and all prior marketplace pages (June 9 — July 17, 2026). The headline find: SeanJ1ang/design-judge-skills (25★) — a production-grade skill pack that decomposes design award submissions into 5 boundary-clear, evidence-driven agent skills with bilingual (Chinese/English) documentation.
New Repos at a Glance¶
| # | Repo | Stars | Skills | Category | Hermes-Ready |
|---|---|---|---|---|---|
| 1 | design-judge-skills | 25 | 5 | Design / Awards | ✅ Tagged |
| 2 | loremaster | 1 | 11 | Project Management | ✅ Tagged |
| 3 | agent-skill-stack | 0 | 1+ | Meta / Skill Discovery | ✅ Tagged |
| 4 | MongrelDB-Hermes | 0 | — | Memory / Infrastructure | ✅ Tagged |
Category Breakdown¶
Design & Awards — design-judge-skills (25★, 6 forks) ⭐ Setup Guide Available¶
Repo: SeanJ1ang/design-judge-skills
Install: npx skills add SeanJ1ang/design-judge-skills
Language: Python · License: Apache 2.0
The first comprehensive agent skill pack for the full design award lifecycle. Five modular skills cover research, evaluation, matching, entry preparation, and submission readiness — each with evidence-driven validation and transparent scoring. Configured for 11 major design awards: iF, Red Dot (Product + Concept), IDEA, DIA, K-Design, GOOD DESIGN Japan, Core77, James Dyson, and EPDA.
The 5 skills:
| Skill | What It Does |
|---|---|
| design-award-search | Retrieve and verify winning entries in the same category from official sources |
| design-evaluation | Score design quality and presentation using evidence-based rubrics |
| design-award-match | Compare awards, tracks, and categories against project fit |
| design-information-prep | Prepare submission text constrained to source materials with word-count validation |
| design-submission-check | Final readiness check against current official rules — go / conditional go / no-go |
Design principles: - Primary sources first — official award pages over third-party summaries - Facts vs. inference separated — user materials, model inferences, and confirm-required items labeled distinctly - Transparent scoring — fit scores aid decisions but don't simulate judging panels - Modular — each skill operates independently with well-defined boundaries
Why it matters for Hermes: The design award submission process involves research, evaluation, writing, and compliance checking — all tasks where agent skills add leverage. These skills work across Claude Code, Codex, OpenClaw, OpenCode, and Hermes Agent. For CorpusIQ's content pipeline, the evaluation framework (evidence-based rubrics, source-first verification) is directly reusable for product review content.
Setup Guide: Design Judge Skills — Full Setup Guide
# Install all 5 skills + shared support package
npx skills add SeanJ1ang/design-judge-skills --skill design-award-search -g -y
npx skills add SeanJ1ang/design-judge-skills --skill design-evaluation -g -y
npx skills add SeanJ1ang/design-judge-skills --skill design-award-match -g -y
npx skills add SeanJ1ang/design-judge-skills --skill design-information-prep -g -y
npx skills add SeanJ1ang/design-judge-skills --skill design-submission-check -g -y
Project Management — loremaster (1★)¶
Repo: loremaster-ai/loremaster
Install: npx skills add loremaster-ai/loremaster
An AI scrum-master / PM skill pack purpose-built for Hermes Agent. Eleven skills that run a team's planning loop around a "lore vault" — a per-project wiki repo that stays the single source of truth. Extracted from a production system running Slack + Jira + GitHub.
Core constraints (by design): - Detect, don't decide — AI surfaces conflicts and drift; humans own judgment - Approval gates — Every write to external systems requires explicit per-item human approval - No invention — Nothing enters the wiki without traceable sources; blanks flagged, never filled - One-way derivation — raw/ → wiki/ → knowledge graph; humans never edit derived layers - Channel = project — Memory, prompts, and graph isolated per project
Architecture: Slack planning loop → Lore vault (git repo with raw/, wiki/, graphify-out/) → Building loop (Claude Code, Cursor) → Jira. Obsidian for human browsing.
Why it matters for Hermes: For teams running Hermes Agent in production with multiple projects, loremaster provides governance structure — human-in-the-loop, source-traceable wiki, per-project isolation. Relevant for CorpusIQ's own multi-project management and potentially for operators managing their own Hermes deployments.
npx skills add loremaster-ai/loremaster -g -y
Meta / Skill Discovery — agent-skill-stack (0★)¶
Repo: neilchen2000-pixel/agent-skill-stack
Install: npx skills add neilchen2000-pixel/agent-skill-stack --skill agent-skill-stack -g -y
License: MIT
A meta-tool that turns natural-language goals into minimal, audited AI agent skill stacks. Discovers, ranks, conflict-checks, and routes skills from GitHub, skills.sh, OpenCLI, and local environments. Derives a workflow from the outcome, reuses local capabilities, discovers direct and indirect helper skills, compares real-world adoption, checks safety and conflicts.
Key features: - Goal → workflow derivation - Multi-source skill discovery (GitHub, skills.sh, OpenCLI, local) - Conflict detection between overlapping skills - Adoption-based ranking - Safety auditing
Why it matters for Hermes: As the skills ecosystem grows past 9,500+ skills, discovery becomes the bottleneck. agent-skill-stack is a discovery layer — useful for Hermes users who need to assemble skill stacks from the broader ecosystem without manually searching.
Memory / Infrastructure — MongrelDB-Hermes (0★)¶
Repo: visorcraft/MongrelDB-Hermes Install: Manual (Python package — not npx skills compatible)
A hybrid long-term memory backend for Hermes Agent using MongrelDB. Combines dense ANN (vector search), sparse retrieval, FM-index, bitmaps, learned ranges, and MinHash for multi-modal retrieval. An alternative to GBrain's PGLite backend for deployments that need stronger retrieval capabilities.
Why it matters for Hermes: For Hermes deployments needing production-grade memory beyond PGLite's single-connection limit. MongrelDB supports multi-process access, making it suitable for multi-agent setups. Early stage (0 stars, July 18 creation) — worth monitoring as an alternative memory backend.
Setup Guides Added¶
This sweep produced one new setup guide: - Design Judge Skills Setup Guide — 5 skills, design award workflow, evidence-driven evaluation framework
Discovery Method¶
GitHub API search for repos tagged hermes-agent and agent-skills created July 17-18, 2026. 40 repos surfaced across two queries. Cross-referenced against all 324 existing catalog entries and 15 prior marketplace discovery pages (June 9 — July 17, 2026) via grep -rl across the full hermes/skills/ directory. 4 new repos confirmed absent from prior sweeps.
Additional repos found (deployment/configuration tools, not skill repos — excluded): - hermes-companion (Android voice, albertosena) - hermes-search-survival-guide (SearXNG config, wu1chenghui) - renz-launcher (Multi-agent launcher, BBRenxo) - Various VPS/Cloudflare/Railway deployment templates
← July 17 Evening | Marketplace Home → Powered by CorpusIQ