Meridian Plugin Hermes Scrub — Development skill for Claude Code

@rynfar/meridian-plugin-hermes-scrub

A [Meridian](https://github.com/rynfar/meridian) plugin that strips [Hermes Agent](https://hermes-agent.nousresearch.com)'s coding-harness fingerprint from the system prompt before it reaches Claude.

Why

Hermes Agent (by Nous Research) can talk to any Anthropic-compatible endpoint via `api_mode: anthropic_messages`, so you can point it straight at Meridian and use your Claude Max subscription. But Hermes injects a built-in `# Finishing the job` framework block into its system prompt — finish-the-job insistence, persistent-memory instructions, and skill-management directives.

Individually those lines are harmless. **Cumulatively the block reads as an autonomous coding-agent harness**, and Anthropic's usage metering flags that signal: the request gets billed as **Extra Usage** instead of under your Max subscription. Once your Extra Usage is depleted, every Hermes request fails with:

400 invalid_request_error: You're out of extra usage. Add more at claude.ai/settings/usage and keep going.

This plugin surgically removes that one block. The request drops back under the metering threshold and is billed under Max (verified empirically — and the fix is independent of the `codeSystemPrompt` preset). Everything else in Hermes' prompt is preserved verbatim:

• the user-editable persona (# Hermes Agent Persona, "You run on Hermes Agent…")
• mid-turn user steering instructions
• the full list
• tools, guidelines, and any user- or harness-appended content

The scrub is **content-scoped** (it runs on every adapter and self-scopes by matching Hermes' fingerprint) and **idempotent** (running it twice is a no-op). It's a safe pass-through for non-Hermes prompts — Claude Code, OpenCode, pi, etc. are left unchanged.

Install

Option 1: Local clone (recommended for dev)

git clone https://github.com/rynfar/meridian-plugin-hermes-scrub.git ~/repos/meridian-plugin-hermes-scrub
cd ~/re