Meridian Plugin Pi Scrub — Development skill for Claude Code
Meridian plugin: strip pi-coding-agent identity fingerprints from the system prompt before it reaches Claude.
How to install Meridian Plugin Pi Scrub
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open rynfar/meridian-plugin-pi-scrub and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Meridian Plugin Pi Scrub does
Meridian plugin: strip pi-coding-agent identity fingerprints from the system prompt before it reaches Claude.
Alternatives in Development
- Discipline — Toggle discipline mode — auto-invoke verification, brainstorming-before-coding, and review checks 2.8k ★
- Projectmem — Open-source coding agent memory 766 ★
- Hetzner VPS Setup Guide — This file is a Claude Code setup prompt 159 ★
README
@rynfar/meridian-plugin-pi-scrub
A [Meridian](https://github.com/rynfar/meridian) plugin that strips pi-coding-agent's identifying fingerprints from the system prompt before it reaches Claude.
Why
Pi's built-in default prompt opens with `"You are an expert coding assistant operating inside pi, a coding agent harness..."` and includes a multi-line "Pi documentation" block that references pi's SDK, extensions, themes, etc.
When Meridian routes pi → Claude Max with `codeSystemPrompt: true`, those pi-identity sections are:
- Redundant — Claude Code's preset already owns identity, tone, safety, and general tool-use guidance.
- A detection fingerprint — they give Anthropic's detection a clear "this isn't Claude Code" signal that can trigger Extra-Usage billing or rate-limit flags.
This plugin surgically removes those two blocks and replaces the identity paragraph with a neutral, generic coding-assistant framing. Everything else in pi's prompt (tools list, guidelines, date, cwd, any user- or harness-appended content, subagent-specific prompt body, project context, skills block) is preserved verbatim.
The scrub is **idempotent** — running it twice on the same string is a no-op.
Install
Option 1: Local clone (recommended for dev)
git clone https://github.com/rynfar/meridian-plugin-pi-scrub.git ~/repos/meridian-plugin-pi-scrub
cd ~/repos/meridian-plugin-pi-scrub
npm install
npm run build
Then point Meridian's plugin config at the built file:
mkdir -p ~/.config/meridian
cat > ~/.config/meridian/plugins.json <<'JSON'
{
"plugins": [
{ "path": "/Users/YOU/repos/meridian-plugin-pi-scrub/dist/index.js", "enabled": true }
]
}
JSON
Restart Meridian (or `curl -X POST http://localhost:3456/plugins/reload`).
Verify at `http://localhost:3456/plugins` — you should see `pi-scrub` listed as **active**.
Option 2: Drop-in file
If you prefer auto-discovery, symlink or copy `dist/index.js` into `~/.config/meridian/plugins
Related Skills
Meridian Plugin Opencode Scrub
Meridian plugin: strip OpenCode identifying fingerprints from the system prompt
Meridian Plugin Hermes Scrub
Meridian plugin: strip Hermes Agent's coding-harness fingerprint from the system prompt to avoid Extra-Usage m
Meridian Plugin Openclaw Scrub
Meridian plugin: stop OpenClaw traffic being metered as a third-party app by removing the prompt sections that
Reverie
Memory for coding agents. Reverie keeps the engineering conclusions your agent reaches, reviewed by you, groun
Spillage
Find the API keys your coding agents spilled into their logs. Claude Code, Codex, Gemini CLI, Cline and more.
Clean Instantiate
Clean a freshly-cloned client repo of canon operator/trust identity BEFORE /ecosystem-init (clear-then-bootstr
Related Agents
Mops Brand QA Gate
The terminal check on every {{client.name}} artifact before it reaches {{paths.approved}} or a human reviewer.
AWS Network Exposure
Determine which AWS network paths from the internet complete end to end, and what the identity attached to eac
Persona Reviewer
Read-only dispatch target for team personas in review, analysis, or assessment mode. Structurally lacks Edit/W