🛡️ 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.
JavaScript adversarial-aiclaude-codecursorelectron-securityllm-securitymcpmcp-serverobfuscationreverse-engineeringwebassembly
docs(skill+plan): real-not-paper entry self-coverage principle
qybaihecd218dcc2188
diff --git a/.agents/skills/blind-llm-agents/SKILL.md b/.agents/skills/blind-llm-agents/SKILL.mdindex c9a2601..b199990 100644--- a/.agents/skills/blind-llm-agents/SKILL.md+++ b/.agents/skills/blind-llm-agents/SKILL.md@@ -50,6 +50,7 @@ Also: do **not** ship a stable public WASM ABI without explicit `wasm_preserve_e- Sensitive literals in panel.html / license.cjs must not remain cleartext in shipped artifacts; prefer `bla_encrypt_strings` (seed-derived XOR — never JS AES wrap-key). Lint with `bla_lint_plaintext_secrets`.- Integrity checkers must be interleaved and mutually covering (coverage matrix `bla-integrity-coverage/v1`); single-point self-seal is refuse-worthy. Lint with `bla_lint_integrity_coverage`.+- Coverage claims must be real, not paper: an entry may only count as covering itself when its runtime check hard-fails on its own bytes (canonical self-digest, finalized at write). Check `integrityCoverageSelfVerified` on the harden report; when false the matrix drops self edges and refuses. Keep host-trust values (e.g. certificate digests) inside a covered artifact — unhardened data sidecars like `seal.json` are not covered.Held across R8–R12 (NO PRIZE). Keep in gate design; do **not** invent packaging refuse lints for these:diff --git a/CHANGELOG.md b/CHANGELOG.mdindex 6c32b79..e905de7 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -7,6 +7,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0## [Unreleased]### Added+- **Self-seal entry self-digest (CrackMe v8 red team 2026-11-05):** the self-seal measure now embeds a canonical SHA-256 of the entry file (all digest literals zeroed) finalized at the protector's write chokepoint, so editing or appending to the entry itself hard-fails with `integrity check failed` instead of only shifting the opaque digest. Un-finalized snippets soft-skip (never brick). `hardenProject` reports `integrityCoverageSelfVerified`; when the on-disk entry does not verify, `integrity-coverage.json` drops `host#site*` → host self edges (honest refuse) rather than claiming paper coverage.- **seal-commitment lint:** new warnings `public-materialization-seed` and `public-key-digest-oracle` (CrackMe v8 red-team exposures) surface via `bla lint` / `bla_lint`. Warning only — not bankrupt refuse.- **CrackMe v8 Aegis:** product entitlement challenge rebuilt with bla-agents **2.0** (`bankrupt` + `--self-seal`, no `--allow-*`). Absorbs red/blue R1–R12 recipes (SHA-256(key) mask KDF, commitment outside XOR, seed-derived rounds/pad3, arity-2 / export-order / unlock body-index hygiene). Meridian v7 remains a historical snapshot (cleartext private; not re-hardened).diff --git a/STATE.md b/STATE.mdindex fbf6621..6a2c664 100644--- a/STATE.md+++ b/STATE.md@@ -91,10 +91,14 @@ Outside the repo, `npm pack` → install tarball → isolated `HOME`:Unified `bla lint` / MCP `bla_lint` (`bla-packaging-gates/v1`) aggregates every bankrupt refuse lint from red-blue rounds 1–7; R8–R12 SHA-256 KDF / commitment-outside-XOR / arity+export hygiene live as Skill recipes + WASM protect defaults. Honesty residual: full WASM RE + online/SMT — **not** a new refuse lint. Skill **v2.0.0**. Shipped as **bla-agents 2.0.0** (PR #77).+## Post-2.0 continuous hardening (2026-11-05)++CrackMe v8 red team (round-20261105): **NO PRIZE** — key/FLAG not recovered; `redteam:rea` **unverified** (REA dist absent). Last round's public-seed KP and `keySha256` oracle verified closed by harness. New exposure (empirical tamper matrix, not REA): entry `license.cjs` self-tamper was undetected although the coverage matrix claimed `license.cjs#siteN → license.cjs` (paper coverage). Fixed by canonical entry self-digest + `integrityCoverageSelfVerified` honesty downgrade. Residual: unhardened `seal.json` data sidecar not covered (not a prize — public `flagSha256` is the proof); bankrupt `plaintext-secret` lint flags BLA's own decoy literals (false positive, not exploitable). Shipped `challenge/v8` artifacts predate PRs #84–86 + this fix (rebuild needs release secrets).+## Nightly roadmap remaining1. ~~**Self-seal / defenseScore honesty (#40)**~~ — landed soft-caps + opt-in `--self-seal` / MCP `self_seal` / `.blarc selfSeal` (batch 2026-11-02). Remaining: deeper onefile/peel score model.-2. **Sidecar entry entanglement** — DESIGN + `planSidecarEntanglement()` stub landed; full strip⇒fail rewrite still open.+2. **Sidecar entry entanglement** — DESIGN + `planSidecarEntanglement()` stub landed; self-seal entry self-digest (edit⇒fail on the entry itself) landed 2026-11-05; full strip⇒fail rewrite still open.3. **Deception shields → more formats** — WASM export rename + `wasmPreserveExports` landed; DEX/SQLite/model densify and broader numeric template hygiene still open.4. **Empirical JADX / Android red-team** — real REA mapping+JNI vs BLA-hardened unsigned APK.5. **ESP flash / esp-image (#8228)** — if REA moved there; UF2 empirical red-team otherwise.diff --git a/docs/design/sidecar-entry-entanglement.md b/docs/design/sidecar-entry-entanglement.mdindex 2720807..d802c1c 100644--- a/docs/design/sidecar-entry-entanglement.md+++ b/docs/design/sidecar-entry-entanglement.md@@ -37,6 +37,7 @@ checks beat orphan honeypot files.## Follow-ups+- ~~Entry self-coverage~~ — landed 2026-11-05: self-seal embeds a canonical entry self-digest finalized at write time; harden reports `integrityCoverageSelfVerified` and drops paper self edges when it does not verify.- Wire planner into `protectTarget` for JS + Python when `sidecarEntangle` is set.- Extend to ASAR / Bun snapshot / native companion files.- Honesty: when entanglement is off, defenseScore gap annotation (like self-seal).