技能生命週期

這是什麼

起草、測試、使用和優化技能的循環。

白話解釋

想像一位園丁的日記——第一次在花圃中發現不熟悉的害蟲時開始記錄，起初是粗糙的筆記，在接下來幾個季節害蟲再次出現時不斷完善，最後當那種害蟲不再出現時就不再翻開。Hermes 的技能經歷相同的軌跡：被寫下，透過重複使用而變得精良，然後在不再有人需要時悄然退役。

為什麼重要

為一個用途起草的技能，如果不經歷其生命週期的其餘階段——在真實任務中測試、反覆使用、然後在實際使用中暴露缺點後進行優化——很少能永遠保持正確。跳過優化步驟，技能腐化就會悄然發生：原本有效的程序不再符合 agent 實際使用的方式，直到失敗前都沒有明顯信號。自我改善技能本質上就是讓這個生命週期持續運行，而不是一次性撰寫就結束。

How it works

Skills the agent creates for itself carry usage telemetry — view count, use count, patch count, and timestamps — and move through active, stale, and archived states based on how long they sit unused. A background process called the curator (agent/curator.py) runs this check on an inactivity trigger rather than a fixed schedule: after 30 days unused a skill becomes stale, after 90 days it’s archived into a recoverable folder rather than deleted. Pinned skills are exempt from every automatic transition. The refining half of the loop runs through the same skill_manage tool (tools/skill_manager_tool.py) that created the skill in the first place, using its patch action to fold in fixes as gaps surface.

A concrete example

Hermes solves a gnarly Kubernetes rollout problem for you after a few failed attempts, then saves the working steps as a new skill.

1. You reuse that skill successfully several times over the next month, and its usage count climbs.
2. Six months pass without another Kubernetes problem, and the curator marks the skill stale.
3. A colleague eventually hits a similar issue, but the skill’s steps assume an older CLI version; they correct it mid-task.
4. Hermes applies that correction as a patch to the skill file rather than leaving the fix as one-off conversation memory.