Other people's setups

01 / What it was

In mid-September we asked for a check of our Hermes setup against the stock one, and looked at how other builders in the community ran their agents. Three open-source projects by another builder stood out: a catalogue of agent personas with skill bundles, a project-stewardship layer that gives a team of agent profiles ownership of a project over time, and a skill that runs a decision past several models.

Rather than pick one, we set up separate test profiles, each with the same skills as the main one, to try them side by side.

02 / How it worked

• Profile 1: the Kanban task board plus a skill that pushes updates back into the chat you are using.
• Profile 2: built-in subagent delegation, routed to specialist profiles with their own skills and tools.
• Profile 3: the stewardship and persona projects together.
• Profile 4: the multi-model “LLM council” skill, for planning.

03 / What we saw

The pain behind it was updates: the task board stopped reporting back, so we had to keep asking the orchestrator for news.

04 / Where it led

Profile 2's routing idea became our profile-pool fork.

Sources

1. Session transcript (the builder's messages). session transcript, 11 Sep 2026 please analyse my hermes setup vs the stock hermes - to see if i am doing things that could be hurting my hermes performance and results
2. Session transcript. session transcript, 11 Sep 2026 it stopped calling back to the orchestrator, so I had to keep messaging the orchestrator for updates again and again
3. Session transcript. session transcript, 11 Sep 2026 they all seem better than my current, messy setup
4. Agent Personas README (third-party project). README of the Agent Personas project A catalogue of 23 personas, 24 specialised skill bundles and 117 reusable skills for building dedicated agents, sub-agents and bots.