The problem is familiar to anyone who has relied on an LLM for regular research: the output changes shape every time. One run hands back a numbered list with citations; the next skips the numbers and buries the links. The model isn\u2019t the culprit \u2014 inconsistent instructions are. Without a stable reference, the agent reinvents its approach on every request. That variability makes delegation impossible. You can\u2019t trust it to run unsupervised.<\/p>\n
This case study documents one real solution: building a reusable skill for an open-source agent called Hermes. The skill automates a weekly research brief \u2014 the kind of task that usually eats an hour of manual prompting and editing. The goal was to produce a consistent, citable document every weekday at 6 a.m. with zero human intervention.<\/p>\n
Before the skill existed, asking the agent to research a topic produced a different result each time. The first attempt might include sources inline; the second would drop them. The structure would shift \u2014 each run re-planned the task from scratch. That\u2019s fine for ad\u2011hoc questions, but for a recurring report, it means re\u2011reading every output to catch omissions.<\/p>\n
The root cause wasn\u2019t the model. It was the absence of a stored procedure. Every request carried the entire context in the prompt, but that context died when the session ended. No reference file told the agent what \u201cdone\u201d looked like.<\/p>\n
The fix requires three elements: an agent identity file, a machine to host it, and a skill file that defines the repeatable procedure. We used Hermes (MIT\u2011licensed, ~236k GitHub stars) on a Hostinger KVM\u20112 plan with 2 cores and 8 GB RAM. The skill was authored using a free web tool called Ordain<\/a><\/strong> (ordain.host), which writes both the identity file and the skill file to the agent\u2019s native format.<\/p>\nStep 1: Define the Agent\u2019s Identity (soul.md)<\/h3>\n