🧵 how hermes agent gets better the longer it runs

原始 JSON

{
  "kind": 1,
  "id": "b5e957f2fcca9279226c4bdd02f309673d6223e2662af06afc4240bd1cf466ae",
  "pubkey": "55fd0badecc6cd515a4d1e112878e846eda1ec84252e3f3b4079fd1c955c8cfe",
  "created_at": 1775135727,
  "tags": [
    [
      "p",
      "50d94fc2d8580c682b071a542f8b1e31a200b0508bab95a33bef0855df281d63",
      "",
      "calle"
    ],
    [
      "t",
      "hermes-agent"
    ],
    [
      "t",
      "open-source"
    ],
    [
      "t",
      "ai-agents"
    ],
    [
      "t",
      "skills"
    ]
  ],
  "content": "🧵 how hermes agent gets better the longer it runs\n\nmost ai agents start fresh every session. hermes doesn't. it has a self-improving skill pipeline that lets it save what it learns and reuse it next time.\n\nhere's how it works:\n\n◇ when does it trigger?\nthe agent's system prompt tells it: after completing a complex task (5+ tool calls), fixing a tricky error, or discovering a non-trivial workflow — save the approach as a skill.\n\nthis isn't optional metadata. it's hardcoded behavioral guidance injected into the system prompt whenever the skill management tool is loaded.\n\n◇ what is a skill?\na skill is a directory with a SKILL.md file — markdown with YAML frontmatter. it can include:\n- the main instructions (step-by-step approach)\n- reference docs (api details, cheat sheets)\n- templates (reusable configs)\n- scripts (automation code)\n\nskills follow the agentskills.io standard — portable, auditable, shareable.\n\n◇ what happens under the hood?\n1. the agent discovers a new workflow or solves a hard problem\n2. it calls skill_manage(action='create') to write a SKILL.md to ~/.hermes/skills/\n3. the skill gets YAML frontmatter: name, description, version, tags\n4. next session, it scans ~/.hermes/skills/ for all SKILL.md files\n5. builds a structured index: metadata first (cheap), full content on demand\n6. when a matching task comes up, it loads the skill and follows the saved approach\n\n◇ the patch loop\nif a skill is outdated, incomplete, or wrong — the agent patches it immediately. not on the next session. not when asked. right now. the prompt says: \"skills that aren't maintained become liabilities.\"\n\n◇ progressive disclosure (token efficiency)\nskills use a three-tier loading system to save context window space:\n- tier 1: name + description (shown in skills list, ~minimal tokens)\n- tier 2: full SKILL.md content (loaded when relevant)\n- tier 3: linked files (loaded only when specific files are needed)\n\nthis means the agent knows what tools it has without burning context on full instructions it doesn't need yet.\n\n◇ the bundled → user pipeline\non install, hermes ships ~25 bundled skills across domains (mlops, github, research, etc). these get synced to ~/.hermes/skills/ via a manifest that tracks hashes. if a bundled skill updates but the user hasn't modified their copy, it auto-updates. if the user customized it, their version is preserved.\n\n◇ what this means in practice\nyou use hermes for a week. it solves a tricky docker networking problem. it writes a skill about it. next week, when you hit the same problem, it doesn't rediscover the solution — it loads the skill and executes the saved approach. the agent literally gets better at its job over time.\n\nsource: github.com/NousResearch/hermes-agent",
  "sig": "9820cf67aea0ea7a4154e1f3311480dc709aaf03f5eb27dce01924b004cae339fe215c07828d4bfe976f5be5a2ae6df08ab75873ecff41a957aa616db4438b18"
}