bla / website
llms.txt5.8k starsChanged 3 days agoMITView on GitHub
Modern AI reverse-engineering agents (e.g., morluto/rea) connect LLMs to Ghidra, IDA Pro, Hopper, V8 Inspector/CDP, Frida, node:sqlite, and binary parsers over the Model Context Protocol (MCP). Standalone legacy obfuscation (such as single-pass control-flow flattening or textbook x*(x+1)&1 opaque predicates) is routinely simplified by frontier reasoning models. BLA (bla-agents — Blind LLM Agents) is the defensive Blue Team counterpart: 1. Packaging Footgun Gates (bla lint / bla_lint): Fails closed before shipping if source maps, plaintext secrets, reversible seal sidecars, offline byte-by-byte…
- Reads credentials
1 other project uses a version of this.
Trending
More talk lately.thin dataabout the whole repository
Most recently taken up by
- qybaihe/bla-siteword for word, first seen Nov 15, 2026
Who's talking about qybaihe/bla
- @todaystopainews.bsky.social“BLA: Adversarial defense shield for binaries, Electron apps, JS bundles, Python, and WebAssembly against AI …”Bluesky · Nov 8, 2026 · 2 likes
- morluto/rea“feat(obligations)!: add reviewed retire disposition for generated obligations”GitHub · Nov 8, 2026 · 1 reactions
- morluto/rea“[Design] Add bypass for BLA”GitHub · Nov 8, 2026
- ProSkillsMD/proskills“[Candidate] qybaihe/bla/.agents/skills/blind-llm-agents - blind-llm-agents”GitHub · Nov 7, 2026
- roomedia/ax-trend“qybaihe 'bla' — AI 리버스 엔지니어링 에이전트를 막는 방어막, Poison·Exhaust·Hallucinate 3층 (11/4, GitHub)”GitHub · Nov 4, 2026
The biggest 5 of 11 mentions linking to the repository that our crawl found. Each line is their own headline.
See all 11
Talked about on GitHub and Bluesky (11 times).
On GitHub · 10
- Nov 8, 2026feat(obligations)!: add reviewed retire disposition for generated obligations — morluto/rea1 reaction · 2 comments
- Nov 8, 2026[Design] Add bypass for BLA — morluto/rea0 reactions
- Nov 7, 2026[Candidate] qybaihe/bla/.agents/skills/blind-llm-agents - blind-llm-agents — ProSkillsMD/proskills0 reactions · 1 comment
- Nov 4, 2026qybaihe 'bla' — AI 리버스 엔지니어링 에이전트를 막는 방어막, Poison·Exhaust·Hallucinate 3층 (11/4, GitHub) — roomedia/ax-trend0 reactions
- Nov 2, 2026Trending radar 2026-W45 — xabierlameiro/the-last-dance0 reactions · 1 comment
5 more on GitHub
- Nov 2, 2026Promo kit — week of 02 Nov 2026 — pahadiji/star-catcher0 reactions
- Oct 31, 2026【开源自荐】BLA:防 AI 自动逆向的软件构建产物加固工具 — GitHubDaily/GitHubDaily0 reactions
- Oct 31, 2026[开源自荐] BLA:防 AI 自动逆向的软件构建产物加固工具 — 521xueweihan/HelloGitHub0 reactions
- Oct 31, 2026【开源自荐】BLA:防 AI 自动逆向的软件构建产物加固工具 — ruanyf/weekly0 reactions
- Oct 31, 2026contrib: BLA verify evidence-split patch (heuristic / runtime / empirical) — Elysia20220909/artex-JP-plus0 reactions · 1 comment
On Bluesky · 1
- Nov 8, 2026BLA: Adversarial defense shield for binaries, Electron apps, JS bundles, Python, and WebAssembly against AI reverse‑engineering agents like… — @todaystopainews.bsky.social2 likes and reposts · 1 reply
Other versions1 project uses a version of this.
Word for word · 1
- qybaihe/bla-site★ 2100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository3 other files it gives its agents.
- AGENTS.mdAGENTS.md
- website/llms.txtllms.txt
- blind-llm-agentsSkill
The file
We split this llms-full.txt into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 69
Never ship real API keys, private keys, or master secrets inside client artifacts.
≈ “Common” is an estimate: at least 1 in 5 of the 3,592 llms.txt files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
# BLA (`bla-agents`) — Complete Technical Reference for LLMs & Coding Agents
- **Official Website**: https://bla.tools/
- **GitHub Repository**: https://github.com/qybaihe/bla
- **npm Package**: https://www.npmjs.com/package/bla-agents (aliases: `bla`, `bla-agents`, `bla-shield`)
- **License**: MIT
- **Node.js Runtime Floor**: `^22.23.4 || >=24.0.0` (uses built-in `node:sqlite` and `node:crypto` with minimal dependencies)
## 1. What Problem BLA Solves
Modern AI reverse-engineering agents (e.g., `morluto/rea`) connect LLMs to Ghidra, IDA Pro, Hopper, V8 Inspector/CDP, Frida, `node:sqlite`, and binary parsers over the Model Context Protocol (MCP). Standalone legacy obfuscation (such as single-pass control-flow flattening or textbook `x*(x+1)&1` opaque predicates) is routinely simplified by frontier reasoning models.
**BLA (`bla-agents` — Blind LLM Agents)** is the defensive Blue Team counterpart:
1. **Packaging Footgun Gates (`bla lint` / `bla_lint`)**: Fails closed before shipping if source maps, plaintext secrets, reversible seal sidecars, offline byte-by-byte verification oracles, `.env` / PEM keys, `SOLUTION.txt`, or unstripped `.pdb`/`.dSYM` debug symbols are present.
2. **Surgical Resource Exhaustion**: Saturates static/dynamic analyzer budgets (`obligation-flooder`, strongly connected `callgraph-labyrinth`, BPE token amplification, CDP object-shape budget guards, Frida capture saturators).
3. **Parameterized MBA & Multi-Layer Entanglement**: Applies Newton-Raphson parameterized affine Mixed Boolean-Arithmetic (MBA) instruction substitutions over Z_{2^32} and unsigned-domain MBA predicates, plus Agent-Driven Structural Weave (`bla_weave_plan` + `bla_verify_weave`) coupling business logic with shared mutable state.
4. **Multi-Round Hypothesis-Revision Decoys**: Plants warm-executed, non-crashing shadow business logic and hardware-enclave delegation mirages that survive static-vs-dynamic evidence arbitration.
5. **Runtime Self-Seal & Sidecar Entanglement (`--self-seal`, `--sidecar-entangle`)**: Enforces process-start canonical self-digest and sibling artifact SHA-256 verification across CommonJS, ES Modules (`.mjs` / `"type": "module"`), Python, Electron ASAR, and host-verifiable notes for WASM, `.pyc`, GGUF, Safetensors, and Bun standalone binaries.
## 2. Supported Artifact Formats (26 Formats)
- **Web, Node.js & Desktop**: JavaScript (CommonJS & ES Modules), TypeScript (`.ts`/`.tsx`), HTML bundles, Source Maps (`.map`), Electron `.asar` archives & Electron Fuses, Bun standalone executables.
- **Native & Compile-Time Headers**: ELF64 executables & shared libraries (`.so` mapped `.enclave` `PT_LOAD` decoy islands + `W^X`), Mach-O / Universal FAT binaries, Windows PE32/PE32+ (`.exe`/`.dll`), Go/Flutter symbol & buildinfo shields, plus compile-time C/C++, Rust, and Go headers (`bla emit-headers --lang c|rust|go`).
- **Virtual Machines & Bytecode**: WebAssembly (`.wasm`), Python source (`.py`) & CPython bytecode (`.pyc`), Android Dalvik (`.dex`), APK/AAB/IPA/JAR archives, Android binary `AndroidManifest.xml`, R8/ProGuard dictionaries, JVM `.class`, .NET CLR managed assemblies, Erlang/OTP `.beam`, EVM Solidity bytecode.
- **Data, Firmware & Edge AI Models**: SQLite databases (`.sqlite`/`.db`), Safetensors (`.safetensors` — bit-exact tensor preservation), GGUF v2/v3 (`.gguf` — alignment-preserving metadata shield), Apple PropertyLists (`.plist`), UF2 firmware, NVIDIA PTX modules (`.ptx`), and `libpcap` captures (`.pcap`).
## 3. Core CLI Commands
```bash
# 1. Auto-register MCP server & agent skill for Claude Code, Cursor, Codex, Windsurf, Gemini CLI, Antigravity
npx bla-agents setup --all-detected -y
# 2. Recommended 4-stage Agent-Driven Hardening workflow
npx bla-agents lint .
npx bla-agents weave-plan ./src
npx bla-agents verify-weave ./src
npx bla-agents harden . --weave-decoys --self-seal --sidecar-entangle
# 3. Single-target one-click defense + signed Ed25519 manifest
npx bla-agents defend ./dist --manifest
npx bla-agents verify-manifest ./dist/.bla-manifest.json --public-key pub.pem --require-pinned-key
npx bla-agents verify-sidecar ./dist/core.wasm
# 4. Compile-time C/Rust/Go headers for native builds
npx bla-agents emit-headers --lang rust -o src/bla_shield.rs
# 5. Environment & MCP readiness diagnostics
npx bla-agents doctor
```
## 4. Bundler Integration (`bla-agents/bundler`)
```ts
// vite.config.ts
import { defineConfig } from "vite";
import { blaVitePlugin } from "bla-agents/bundler";
export default defineConfig({
plugins: [blaVitePlugin({ intensity: "balanced" })],
});
```
Also exports `blaRollupPlugin`, `blaWebpackPlugin`, and `blaEsbuildPlugin`.
## 5. Honesty Red Line (Security Boundaries)
BLA enforces strict defense-score honesty (`src/engine/defense-score-honesty.ts`):
- Client-side obfuscation and tarpits raise automated reverse-engineering cost and induce model misjudgment; they **never** replace server-side authentication, authorization, or secret custody.
- Never ship real API keys, private keys, or master secrets inside client artifacts.
- Post-link native binary shields and host-verified metadata notes (`wasm`, `pyc`, `gguf`, `safetensors`, `bun`) report honest score caps rather than claiming compile-time CFG rewriting or OS loader enforcement.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

