Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/samtalki/agentrepl.jl/claude-mdgit clone --depth 1 https://github.com/samtalki/AgentREPL.jlWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/samtalki/agentrepl.jl/claude-md)<a href="https://agentmods.dev/instructions/samtalki/agentrepl.jl/claude-md"><img src="https://agentmods.dev/badge/instructions/samtalki/agentrepl.jl/claude-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.03856 | $0.03856 |
| Opus 5 | $0.01928 | $0.01928 |
| Sonnet 5 | $0.00771 | $0.00771 |
| Haiku 4.5 | $0.00386 | $0.00386 |
Grade B, and why
AgentREPL.jl CLAUDE.md scanned grade B with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
- **STDIO transport only**: No network ports for security. The `session attach` socket is the one local IPC channel, a Unix domain socket chmod 600 (owner only) How it starts
The opening of the file, as written. The whole thing — 226 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Project Overview
AgentREPL.jl is a Julia package providing a persistent REPL for AI agents via MCP (Model Context Protocol) STDIO transport. It solves Julia's "Time to First X" (TTFX) problem by maintaining persistent Julia worker subprocesses, avoiding the 1-2 second startup penalty for each command. Supports multiple named sessions and Revise.jl hot-reloading.
Build and Test Commands
# Run all tests (Aqua quality + unit suites; fast, no subprocess spawning)
julia --project=. -e "using Pkg; Pkg.test()"
# Include the MCP protocol integration tests (spawns a server subprocess, ~60s)
AGENTREPL_E2E=true julia --project=. -e "using Pkg; Pkg.test()"
# Run a specific test file directly
julia --project=. test/test_eval.jl
# Start the MCP server manually (for debugging)
julia --project=. -e "using AgentREPL; AgentREPL.start_server()"
# Start server with a specific project activated
JULIA_REPL_PROJECT=/path/to/project julia --project=. bin/julia-repl-server
Architecture
Multi-Session Worker Subprocess Model
AgentREPL uses a multi-session worker subprocess architecture (via Malt.jl):
- The MCP server runs in the main Julia process
- Each named session has its own
Malt.Workerprocess for code evaluation - Sessions are isolated: separate variables, packages, and project environments
resetkills a session's worker and spawns a fresh one (true hard reset)activateswitches a session's active project/environment- Revise.jl is auto-loaded on workers for hot-reloading support
Why Malt, not Distributed.jl: Malt spawns workers with monitor_stdout=false/monitor_stderr=false, so worker output goes to private pipes (drained to the server's stderr by _start_output_drain!) and can never reach the main process's stdout — which is the MCP JSON-RPC transport. Malt also has no global cluster state (each Worker is an independent object) and provides graceful stop (exit → SIGTERM → SIGKILL), isrunning, and TerminatedWorkerException, replacing hand-rolled lifecycle code. IPC is Malt.remote_eval_fetch(Main, w, expr) via the _remote_eval_fetch helper.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 226 lines · 3,856 tokens per session scan B 5aed700d2ac0
AgentREPL.jl CLAUDE.md is an instructions file published in the GitHub repository samtalki/AgentREPL.jl (6 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 3,856 tokens to every session, about $0.0193 per session on Opus 5. A static security scan graded it B with 1 finding (asks for root). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
BestieTemplate.jl AGENTS.md
Instructions for JuliaBesties/BestieTemplate.jl, covering agents.md, architecture, development commands, testing via julia-mcp and adding a new addfeature(:feature).
JuliaFEM.jl AGENTS.md
Instructions for JuliaFEM/JuliaFEM.jl, covering 1. what this project is, 2. where things live, 3. current architecture (0.x), 3.1 dof specification and single field.
intelligent-terminal rust.instructions.md
Concise Rust coding conventions for this repository.
InvestSkill GEMINI.md
Instructions for yennanliu/InvestSkill, covering investskill — gemini cli setup & usage guide, installation & setup, quick start, navigate to the investskill directory and start gemini cli (loads gemini.md automatically).
she-love-me CLAUDE.md
Instructions for 863401402/she-love-me, covering claude.md and 唯一工作流.
superset CLAUDE.md
Claude Code instructions for superset-sh/superset, a project described as: Superset is an agentic IDE to orchestrate 100+ coding agents in parallel. Run any agent with your own subscription.