Repository navigation
Conversation
|
Try this branchThe engine and the scripts resolve independently, so this runs the changed export COMMUNITY_SCRIPTS_URL=https://raw.githubusercontent.com/Estcool9/DevScripts-voicestudio/feat/rea-reverse-engineering-mcp
bash -c "$(curl -fsSL "$COMMUNITY_SCRIPTS_URL/ct/rea.sh")"Both lines are needed. Without The same command works on an Incus host: the engine detects the platform and Useful while testing
export COMMUNITY_SCRIPTS_URL=https://raw.githubusercontent.com/Estcool9/DevScripts-voicestudio/feat/rea-reverse-engineering-mcp
dev_mode=net,keep bash -c "$(curl -fsSL "$COMMUNITY_SCRIPTS_URL/ct/rea.sh")"Against a core branch as wellAdd |
✍️ Description
Adds a Debian 13 unprivileged LXC helper for REA, an MCP-based reverse-engineering toolkit. The installer provisions Node.js 24, Temurin JDK 25, Ghidra, a checksum-verified JADX MCP adapter, and a dedicated unprivileged systemd service with Ghidra/JADX health checks. REA's HTTP MCP listener has no built-in OAuth, so the catalog metadata explicitly requires restricting TCP 8765 to a trusted MCP gateway and terminating OAuth 2.0 at a TLS-enabled MCPHub. The current stack was deployed and provider-checked manually in CT 107 on Proxmox VE; the new helper's complete DevScripts
build_containerflow has not yet been exercised, so this PR is submitted as a draft for end-to-end testing.🔗 Related PR / Issue
Link: #
✅ Prerequisites
🧩 Script
Name: REA
Tool: REA is a reverse-engineering CLI and MCP server for static analysis of binaries, applications, and firmware. It integrates Ghidra and JADX and exposes analysis workflows to AI agents and developer tools.
🗂️ Category
ct/+install/)vm/)tools/addon/tools/pve/🖥️ Platform – tested on
🏗️ Architecture
🛠️ Type of Change
🔍 Code & Security Review
CODE-AUDIT.md&CONTRIBUTING.mdguidelinesAppName.sh,AppName-install.sh,AppName.json)📋 Additional Information
Static checks performed on the three new files:
bash -n ct/rea.sh install/rea-install.shpython3 -m json.tool json/rea.jsongit diff --checkThe live manual deployment has healthy Ghidra and JADX provider diagnostics, and REA is reachable through a private MCPHub route. It is not a substitute for exercising this PR through the DevScripts container builder. The MCP endpoint is intentionally unauthenticated at the application layer; keep it private behind a trusted gateway/firewall.
📦 Application Requirements (for new scripts)
🌐 Source
🤖 AI Assistance
AGENTS.mdand.github/agents/pve-script-creator.agent.mdas guidance, and the output has been reviewed and corrected to match those guidelines.LLM usage: GPT-6 (reasoning level not exposed in the session metadata) was used to draft and revise the helper scripts, metadata, and this PR description. The resulting files were checked against the repository guidance, shell syntax, ShellCheck, JSON validity, and the upstream update-helper conventions. The full DevScripts
build_containertest is still pending.