Reading the repo
Reading the repo
How to install on Linux
🛡️ BLA (Blind LLM Agents) — Adversarial defense shield that protects binaries, Electron apps, JS bundles, Python, and WebAssembly against AI reverse-engineering agents like REA. Poison -> Exhaust -> Hallucinate.
Saved setup from 10 h ago
Will it work on Linux?
docs/installation.mdBLA (bla-agents — Blind LLM Agents) is distributed as a zero-config Node.js CLI and MCP server that runs anywhere Node.js ^22.23.4 || >=24.0.0 is installed (matches package.json engines).
npx)Run BLA directly without installing anything globally:
# 1. Register BLA MCP Server + Skill into Claude Code, Cursor, Codex, Windsurf, Gemini CLI, etc.
npx bla-agents setup
# 2. Check your environment and configured AI clients
npx bla-agents doctor
# 3. Audit an artifact for REA exposure before release
npx bla-agents audit ./dist/app.js
# 4. Protect the artifact with BLA's three-stage shield
npx bla-agents protect ./dist/app.js --mode bankrupt
# 5. Verify functional preservation and REA analyzer degradation
npx bla-agents verify ./dist/app.js ./dist/app.bla.js
npm install --save-dev bla-agents
# or
pnpm add -D bla-agents
# or
bun add -d bla-agents
Then add a post-build protection script to your package.json:
{
"scripts": {
"build": "vite build",
"postbuild": "bla protect ./dist/assets/index.js --in-place --mode bankrupt --emit-sourcemap"
}
}
npm install -g bla-agents
# Both `bla` and `bla-agents` binaries are registered:
bla --version
bla-agents doctor
bla setup)Just like npx reverse-engineer-anything setup, npx bla-agents setup auto-discovers installed AI coding agents on your machine and registers:
bla MCP stdio server (bla_audit, bla_protect, bla_verify, bla_benchmark, bla_emit_headers, bla_doctor)blind-llm-agents skill (SKILL.md + Counter-REA Matrix)| Client ID | Name | Config Location | Skill Location |
| :--- | :--- | :--- | :--- |
| claude-code | Claude Code | ~/.claude.json | ~/.claude/skills/blind-llm-agents |
| cursor | Cursor | ~/.cursor/mcp.json | ~/.cursor/skills/blind-llm-agents |
| codex | OpenAI Codex | ~/.codex/config.toml | ~/.codex/skills/blind-llm-agents |
| windsurf | Windsurf | ~/.codeium/windsurf/mcp_config.json | ~/.codeium/windsurf/skills/blind-llm-agents |
| gemini | Gemini CLI | ~/.gemini/settings.json | ~/.gemini/skills/blind-llm-agents |
| opencode | OpenCode | ~/.config/opencode/opencode.json | ~/.config/opencode/skills/blind-llm-agents |
npx bla-agents setup --client claude-code,cursor
To cleanly remove BLA from all agent configuration files and delete installed skill folders:
npx bla-agents uninstall
If you are shipping a native binary compiled with clang, gcc, rustc, or go build, you can embed BLA's compile-time counters directly into .rodata before linking:
npx bla-agents emit-headers ./include/shield --seed my-release-v1
#include "bla_shield.h" and call BLA_ANCHOR_SHIELD(); inside main().mod bla_shield; and call bla_shield::anchor_shield(); inside fn main().bla_shield.go in your package and call AnchorBlaShield() inside func main().Then run bash setup.sh. It stops at the first error and is safe to run again. Servers and tests aren't started; it lists them at the end. Read it first.
git clone https://github.com/qybaihe/bla.git cd bla
Node.js 24
# needs nvm: https://github.com/nvm-sh/nvm nvm install nvm use
npm ci
npm run dev
npm test
# Node.js ^22.23.4 or 24.x recommended (REA red-team needs zlib.zstd on 22.23+)
npx bla-agents setup # register MCP + skill with your coding agent
npx bla-agents harden . # one-command Counter-REA pass on the current project
npx bla-agents benchmark # 14-target showcase (local estimates + runtime checks)
Ask your agent after setup (skill guides one-click harden):
Harden my software against REA
Or run stages explicitly: audit · protect · verify · defend — see one-click agent defense and CLI and MCP.
No reports for Linux yet.
Opens a chat with this manual already filled in, for Linux. It uses your own account; gitmanual sends nothing.