{"slug": "unlimited-skill", "title": "Unlimited skill", "summary": "Hermes Agent introduces a 5-layer hybrid routing plugin that enables unlimited skills with near-zero latency. The system uses BM25 and trigger phrase matching for O(1) lookup, supports bilingual Chinese and English queries, and includes self-training and drift monitoring capabilities.", "body_md": "Unlimited skill engine — Hermes Agent 5-layer hybrid routing plugin.\n\nLike a maître d' at a fine restaurant: you say \"I'm hungry\", they seat you at the best table — no need to flip through a 400-dish menu.\n\nSee the Difference\n\nMe: \"Help me optimize my website\"\n\n❌ Before: AI confused, spits out irrelevant text\n\n✅ After: AI understands instantly, calls the \"Performance Optimization\" skill\n\nMe: \"Thanks, well said\"\n\n❌ Before: Still suggesting skills like a pushy salesperson\n\n✅ After: A quiet \"you're welcome\" — just chatting normally\n\nArchitecture\n\n代码\n\nUser Query\n\n│\n\n▼\n\n[1] Task Gate (rules + LLM hybrid)\n\n├── Not a task → skip (zero overhead)\n\n└── Task →\n\n│\n\n▼\n\n[2] Exact Trigger Match\n\n├── Hit → Top-3 injected\n\n└── Miss →\n\n│\n\n▼\n\n[3] Trigger Word Overlap (overlap scoring)\n\n├── Hit → Top-3 injected\n\n└── Miss →\n\n│\n\n▼\n\n[4] Levenshtein Fuzzy (typo tolerance)\n\n├── Hit → Top-3 injected\n\n└── Miss →\n\n│\n\n▼\n\n[5] LLM Fallback (edge case catch-all)\n\n→ Top-3 injected\n\nFeatures\n\nUnlimited skills — BM25 + trigger phrase matching, O(1) lookup, skill count doesn't affect latency\n\nBilingual CN/EN — 2,918 trigger phrases covering real Chinese and English user queries\n\nSelf-training loop — user corrections → feedback log → cron auto-training\n\nDrift monitor — runs 366 test cases daily at 4AM, alerts if accuracy drops >5%\n\nZero noise — scores 0.5–1.0 have meaning; no BM25 random guesses\n\n100% local — all indexes on-device, no external APIs\n\nInstall\n\n代码\n\n· bash\n\nhermes skills install [https://github.com/wuykjl/unlimited-skills](https://github.com/wuykjl/unlimited-skills)\n\nInstall and go. Works in Chinese and English. Learns on its own. Monitors itself for drift.\n\nWho's It For?\n\nYou say \"write me a script\", \"deploy this\", \"check for security holes\" — daily\n\n300+ skills installed and you don't know which ones are actually being used\n\nYou don't want to waste time thinking about \"which skill to pick\"\n\nAcknowledgments\n\nThis project's skill routing is built on the open skill ecosystems of:\n\nECC (Enterprise Coding Conventions) — 117-rule coding standard framework\n\nAnthropic Skills — Official Agent Skills repository\n\nClaude Code Plugins — Claude Code official plugin ecosystem\n\nAnd all individual developers who contributed skills to the Hermes ecosystem", "url": "https://wpnews.pro/news/unlimited-skill", "canonical_source": "https://dev.to/wuykjl/unlimited-skill-2dga", "published_at": "2026-06-16 02:42:40+00:00", "updated_at": "2026-06-16 03:17:31.891139+00:00", "lang": "en", "topics": ["ai-agents", "developer-tools", "natural-language-processing", "machine-learning"], "entities": ["Hermes Agent", "BM25", "ECC", "Anthropic Skills", "Claude Code Plugins"], "alternates": {"html": "https://wpnews.pro/news/unlimited-skill", "markdown": "https://wpnews.pro/news/unlimited-skill.md", "text": "https://wpnews.pro/news/unlimited-skill.txt", "jsonld": "https://wpnews.pro/news/unlimited-skill.jsonld"}}