run-agentrepl

A skill for starting and testing AgentREPL.jl, an MCP server that exposes a persistent Julia REPL to an AI coding assistant.

In plain words
What is it for?
Starting the server, running its tests, evaluating Julia code, rendering plots, and confirming behavior through the server.
Why use it?
It provides a defined way to launch the server, send Julia code to it, and check that changes work.

Skill for Claude CodeCodex

Install

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.

agentmods
npx agentmods add skills/samtalki/agentrepl.jl/run-agentrepl
Any agent
npx skills add samtalki/AgentREPL.jl --skill run-agentrepl
Clone the repo
git clone --depth 1 https://github.com/samtalki/AgentREPL.jl

Made for: Claude Code, Codex.

Per session 61 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,933 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00061 $0.01933
Opus 5 $0.00030 $0.00966
Sonnet 5 $0.00012 $0.00387
Haiku 4.5 $0.00006 $0.00193

Measured 2d ago against content hash 758dafdca036, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

run-agentrepl scanned grade A 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 2d ago.

The scan reads SKILL.md. This mod also ships 1 executable file (driver.mjs), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

- **STDIO only, no port.** You cannot `curl` this server. The single local IPC channel is `session attach` (a chmod-600 Unix socket for an interactive human REPL), which needs `tmux` and is out of scope for headless driv
.claude/skills/run-agentrepl/SKILL.md · 116 lines

How it starts

The opening of the file, as written. The whole thing — 116 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AgentREPL.jl is an MCP server that exposes a persistent Julia REPL (eval, sessions, pkg, revise, plots) over JSON-RPC on STDIO. There is no GUI and no network port, so you drive it by being its MCP client: spawn the server, do the initialize handshake, call tools. The driver .claude/skills/run-agentrepl/driver.mjs does exactly that — use it for everything below.

All paths are relative to the repo root (/Users/sam/Research/AgentREPL.jl).

Prerequisites

Verified on macOS (darwin), Julia 1.12.6, Node v26. Linux is equivalent — there is nothing platform-specific in the launch path.

  • Julia 1.12+ — the julia binary on PATH. Installed here via juliaup.
  • Node 18+ — only for driver.mjs. Uses Node stdlib only, no npm install.

Setup

Manifest.toml is gitignored, so resolve and precompile deps once after clone (~30s cold):

julia --project=. -e "using Pkg; Pkg.instantiate(); Pkg.precompile()"

Optional — enable Revise.jl hot-reload. Revise is only a test-extra of this package, not a runtime dep, and the worker inherits your global env on its load path. So hot-reload works only if Revise lives in your default environment:

julia -e 'using Pkg; Pkg.activate(); Pkg.add("Revise")'

Without this, info reports Revise.jl: not available and the server logs a Could not load Revise.jl on worker warning (and records a session note) per worker spawn. Eval is unaffected.

Run (agent path)

Drive the server with driver.mjs. It spawns julia --project=. bin/julia-repl-server, handles the handshake, and skips the non-JSON chatter the server interleaves on its streams.

Full smoke test — launches, exercises all 8 tools, prints PASS/FAIL, exits non-zero on any failure (~10s warm, after precompile):

node .claude/skills/run-agentrepl/driver.mjs

One-shot eval — spawn, eval, print the formatted result, exit:

node .claude/skills/run-agentrepl/driver.mjs eval 'using Statistics; mean([1,2,3,4,5])'

Read the full file on GitHub · 116 lines

Files

What ships with it

1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

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.

  1. 2d ago First seen · 116 lines · 61 tokens per session scan A 758dafdca036

Subscribe to this mod's changes

run-agentrepl is a skill published in the GitHub repository samtalki/AgentREPL.jl (6 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 61 tokens to every session and 1,933 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other skills, from other repositories

harness-creator

Build, audit, and improve harnesses that make AI coding agents reliable: AGENTS.md/CLAUDE.md instruction files, feature/state tracking, verification gates, scope boundaries, session handoff, memory persistence, context budgets, tool-permission safety, and multi-agent coordination. Use this whenever a coding agent is…

walkinglabs/learn-harness-engineering · 142 tokens

hive.worker-delegation

Concrete patterns for breaking colony work into parallel worker jobs via runplaybook — when fan-out helps, how to model the goal as a tracker table, write the worker skill, author the playbook, pilot, and let convergence retry/resume the gap.

aden-hive/hive · 58 tokens

gget

Fast CLI/Python queries to 20+ bioinformatics databases. Use for quick lookups: gene info, BLAST searches, AlphaFold structures, enrichment analysis. Best for interactive exploration, simple queries. For batch processing or advanced BLAST use biopython; for multi-database Python workflows use bioservices.

synthetic-sciences/openscience · 66 tokens

histolab

Lightweight WSI tile extraction and preprocessing. Use for basic slide processing tissue detection, tile extraction, stain normalization for H&E images. Best for simple pipelines, dataset preparation, quick tile-based analysis. For advanced spatial proteomics, multiplexed imaging, or deep learning pipelines use pathml.

synthetic-sciences/openscience · 62 tokens

pydicom

Python library for working with DICOM (Digital Imaging and Communications in Medicine) files. Use this skill when reading, writing, or modifying medical imaging data in DICOM format, extracting pixel data from medical images (CT, MRI, X-ray, ultrasound), anonymizing DICOM files, working with DICOM metadata and tags…

synthetic-sciences/openscience · 110 tokens

scientific-schematics

Create publication-quality scientific diagrams using Nano Banana 2 AI with smart iterative refinement. Uses Gemini 3.1 Pro Preview for quality review. Only regenerates if quality is below threshold for your document type. Specialized in neural network architectures, system diagrams, flowcharts, biological pathways…

xintaofei/codeg · 68 tokens