hermes - 💡(How to fix) Fix Question: is there a convention for how system prompt modules format their headings/sections?

Question: is there a convention for how system prompt modules format their headings/sections?

Description

I was reading through how the system prompt is assembled and noticed something I wanted to ask about — it seems like different modules use different formatting styles, and I wasn't sure if this was intentional or just hasn't been discussed yet.

I'm not deeply familiar with the codebase so I may be missing context, but I figured I'd lay out what I observed and ask.

What I found

All references are from agent/prompt_builder.py and tools/memory_tool.py on main.

A few modules use no heading at all — just plain paragraphs:

Some modules use # (level-1 Markdown heading) — all of these are conditionally injected:

One module uses ## (level-2 heading):

Two modules use ═══ ASCII separators + uppercase labels:

And all modules are joined by "\n\n".join(...) — no wrapping or boundary markers between them.

What the assembled result looks like

In a normal chat session (Kanban off, Computer Use off, skills tools loaded), the system prompt roughly looks like:

You are Hermes Agent, an intelligent AI assistant...        ← bare paragraph
If the user asks about configuring...                        ← bare paragraph
You have persistent memory across sessions...                ← bare paragraph
When the user references something...                        ← bare paragraph
After completing a complex task...                           ← bare paragraph
# Tool-use enforcement                                      ← h1
...
## Skills (mandatory)                                       ← h2
...
[context tier: HERMES.md / AGENTS.md / CLAUDE.md]
══════════════════
MEMORY (your personal notes)
══════════════════
...

Why I'm asking

I was trying to understand the system prompt's structure and found it a bit hard to follow. The first five modules blend together visually, and different modules seem to use different "visual languages" (Markdown headings vs ASCII separators vs bare text).

I've seen other projects handle this with:

• Consistent Markdown heading levels across all modules
• Or XML-style tags wrapping each section (like <guidance>...</guidance>)

But I genuinely don't know if either approach makes sense here, or if there are constraints (prompt caching, token efficiency, etc.) that I'm unaware of.

Just wanted to ask: has there been any discussion about a unified formatting convention for system prompt modules? Or is the current diversity of styles intentional?