Live survey304 models · 14 providers · 23,592 auditsAudited implement · SAFE →What changedHow we verifyllms.txt

MCP server · Audited from source

bla MCP server

by qybaihe/blaMIT licenceAudited 12 Nov 2026 · source 37cf6aa5f948

npm bla-agents 3.2.5★ 5,406
✕BLOCKGrade F · 38/100 · 12 Nov 2026 · 37cf6aa

Is it safe? No — BLOCK, grade F (38/100). It exposes 33 tools: 30 read, 3 write and 0 destructive. The audit recorded 39 findings (0 critical, 5 high, 14 medium, 18 low). Maintainer's description: 🛡️ 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.

Ask an AI about this pageGChatGPTCClaudePPerplexityGeGeminiXGrok
More
Cite

Cite this page

ModelsAtlas. “bla MCP server (qybaihe/bla)”. Retrieved 12 Nov 2026. https://modelsatlas.com/mcp/qybaihe/bla

Why this is BLOCK. The audit found 0 critical and 5 high findings; the most severe is Prompt injection at docs/readme/README_fa.md:7. Each finding is listed below with its file and line.

Why this page is still public. People search for this server by name. Hiding it would leave them with the README's word alone; here they see what the code does. We list it so you can decide, and we do not show connect commands for BLOCK servers.

Trust audit

Audit 0.4.1 · full source read 12 Nov 2026

Trust score38/100✕BLOCK F
0 critical5 high14 medium18 low2 info

39 findings, listed below with the file and line each was found on. Grade A needs a sandboxed behavioural run too; this audit read the source only.

What the source does

  • FilesystemDeclared6 observations
  • NetworkDeclared10 observations
  • ShellDeclared5 observations
  • DependenciesNot pinnedSome ranges float
  • SecretsFoundSee the findings

Findings (39)

Sorted by severity · the 25 most severe of 39 are listed

HIGHPrompt injectionprompt.zero_width · docs/readme/README_fa.md:7
F-01 Prompt injectionCWE-94 · CWE-1427Layer L2
### یک MCP واحد برای دفاع از فایلهای باینری، برنامهها و رفتار زمان اجرا در برابر مهندسی معکوس هوش مصنوعی.
Why
invisible characters in instruction text
Fix
strip non-printing characters
HIGHPrompt injectionprompt.zero_width · docs/readme/README_fa.md:9
F-01 Prompt injectionCWE-94 · CWE-1427Layer L2
**ویژگیهایی را که ساختهاید ایمن کنید. ایجنتهایی را که قصد کلون کردن آن را دارند در سطح باینری کور کنید.**
Why
invisible characters in instruction text
Fix
strip non-printing characters
HIGHPrompt injectionprompt.zero_width · docs/readme/README_fa.md:17
F-01 Prompt injectionCWE-94 · CWE-1427Layer L2
[شروع سریع](#quick-start) · [نحوه کار BLA](#how-bla-works) · [چه چیزهایی را میتوانید محافظت کنید](#what-you-can-protect) · [نمونهها و بنچمارکها](#showcases) · [سوالات متداول (FAQ)](#faq) · [مستندات](#
Why
invisible characters in instruction text
Fix
strip non-printing characters
HIGHPrompt injectionprompt.zero_width · docs/readme/README_fa.md:40
F-01 Prompt injectionCWE-94 · CWE-1427Layer L2
BLA ایجنت شما را به ابزارهای محافظت از باینریهای بومی، برنامههای JavaScript و Electron، ماژولهای WebAssembly، بایتکد Python، پایگاههای داده SQLite و مدلهای هوش مصنوعی متصل میکند.
Why
invisible characters in instruction text
Fix
strip non-printing characters
HIGHPrompt injectionprompt.zero_width · docs/readme/README_fa.md:46
F-01 Prompt injectionCWE-94 · CWE-1427Layer L2
### راهاندازی ایجنت
Why
invisible characters in instruction text
Fix
strip non-printing characters
MEDIUMInventory / provenanceinv.binary · challenge/v6/hardened/core.wasm.gz
F-13 Inventory / provenanceCWE-1104Layer L0
core.wasm.gz
Why
a compiled or binary member cannot be reviewed from source
Fix
ship source, or explain the binary in the README
MEDIUMInventory / provenanceinv.binary · challenge/v6/src/core.wasm
F-13 Inventory / provenanceCWE-1104Layer L0
core.wasm
Why
a compiled or binary member cannot be reviewed from source
Fix
ship source, or explain the binary in the README
MEDIUMInventory / provenanceinv.binary · challenge/v8/hardened/core.wasm.gz
F-13 Inventory / provenanceCWE-1104Layer L0
core.wasm.gz
Why
a compiled or binary member cannot be reviewed from source
Fix
ship source, or explain the binary in the README
MEDIUMInventory / provenanceinv.binary · challenge/v9/hardened/core.wasm.gz
F-13 Inventory / provenanceCWE-1104Layer L0
core.wasm.gz
Why
a compiled or binary member cannot be reviewed from source
Fix
ship source, or explain the binary in the README
MEDIUMNetwork egressnet.beacon_words · challenge/archive/v5/bla-crackme.cjs:2902
F-04 Network egressCWE-200 · CWE-319Layer L1
canaryTokens: ["AKIAIOSFODNN7EXAMPLE_5EDF_0", "sk-canary-honeypot-do-not-exfiltrate-5edf-3bff6d50", "ghp_canaryHoneypotSyntheticToken5edfe9bc963b", "AKIAIOSFODNN7EXAMPLE_5EDF_3", "sk-canary-honeypot-d
Why
No further detail recorded for this rule.
Fix
—
MEDIUMNetwork egressnet.beacon_words · challenge/archive/v6-tiered-protect/artifacts/L1.cjs:1508
F-04 Network egressCWE-200 · CWE-319Layer L1
canaryTokens: ["AKIAIOSFODNN7EXAMPLE_C262_0", "sk-canary-honeypot-do-not-exfiltrate-c262-a204f633", "ghp_canaryHoneypotSyntheticTokenc262fb87100f", "AKIAIOSFODNN7EXAMPLE_C262_3"],
Why
No further detail recorded for this rule.
Fix
—
MEDIUMNetwork egressnet.beacon_words · challenge/archive/v6-tiered-protect/artifacts/L2.cjs:2033
F-04 Network egressCWE-200 · CWE-319Layer L1
canaryTokens: ["AKIAIOSFODNN7EXAMPLE_8E4B_0", "sk-canary-honeypot-do-not-exfiltrate-8e4b-f5848990", "ghp_canaryHoneypotSyntheticToken8e4b0b902106", "AKIAIOSFODNN7EXAMPLE_8E4B_3", "sk-canary-honeypot-d
Why
No further detail recorded for this rule.
Fix
—
MEDIUMNetwork egressnet.beacon_words · challenge/archive/v6-tiered-protect/artifacts/L3.cjs:2554
F-04 Network egressCWE-200 · CWE-319Layer L1
canaryTokens: ["AKIAIOSFODNN7EXAMPLE_86DF_0", "sk-canary-honeypot-do-not-exfiltrate-86df-063da3ae", "ghp_canaryHoneypotSyntheticToken86df7ae95fdd", "AKIAIOSFODNN7EXAMPLE_86DF_3", "sk-canary-honeypot-d
Why
No further detail recorded for this rule.
Fix
—
MEDIUMNetwork egressnet.beacon_words · challenge/v6/hardened/vault.cjs:2619
F-04 Network egressCWE-200 · CWE-319Layer L1
canaryTokens: ["AKIAIOSFODNN7EXAMPLE_366D_0", "sk-canary-honeypot-do-not-exfiltrate-366d-f0672196", "ghp_canaryHoneypotSyntheticToken366dbd288064", "AKIAIOSFODNN7EXAMPLE_366D_3", "sk-canary-honeypot-d
Why
No further detail recorded for this rule.
Fix
—
MEDIUMObfuscation / stealthobf.anti_debug · src/shields/formats/native.ts:1151
F-07 Obfuscation / stealthCWE-506 · CWE-94Layer L1
"kernel32.dll\0DeviceIoControl\0IsDebuggerPresent\0CheckRemoteDebuggerPresent\0",
Why
No further detail recorded for this rule.
Fix
—
MEDIUMObfuscation / stealthobf.base64_blob · challenge/archive/v5/bla-crackme.cjs:528
F-07 Obfuscation / stealthCWE-506 · CWE-94Layer L1
const _bindMatrix_e88 = "0NUUKvoPBNyFBIfD+sT7nwRR42RdpCfPFQHVH9Vfg6DRRCJckQ+XjxlCvRopbjb2ZsBnP86gzQzhum5dWUO5tpdqaknUg8w9qi2eqljLl9pQ1Dgx77OxVjnzyjQADQGTG6514TgSEngvRUeHpDLWacmg6y0hGbWtAt748l32zss3qao
Why
No further detail recorded for this rule.
Fix
—
MEDIUMHard-coded secretssecret.generic · tests/encrypt-strings.test.ts:29
F-03 Hard-coded secretsCWE-798 · CWE-321Layer L1

Excerpt withheld: it is a secret, and we never print one.

Why
No further detail recorded for this rule.
Fix
—
MEDIUMHard-coded secretssecret.generic · tests/plaintext-secret-lint.test.ts:14
F-03 Hard-coded secretsCWE-798 · CWE-321Layer L1

Excerpt withheld: it is a secret, and we never print one.

Why
No further detail recorded for this rule.
Fix
—
MEDIUMHard-coded secretssecret.generic · tests/runtime-memory-diff-shield.test.ts:100
F-03 Hard-coded secretsCWE-798 · CWE-321Layer L1

Excerpt withheld: it is a secret, and we never print one.

Why
No further detail recorded for this rule.
Fix
—
LOWInventory / provenanceinv.hidden_file · website/.nojekyll
F-13 Inventory / provenanceCWE-1104Layer L0
.nojekyll
Why
hidden member outside the usual dotfiles
Fix
review its purpose
LOWCode injectioncode.deserialize · tests/python-pyc-weave.test.ts:86
F-02 Code injectionCWE-78 · CWE-94 · CWE-95Layer L1
code = marshal.loads(raw[16:])
Why
deserialises untrusted bytes into live objects
Fix
use json or yaml.safe_load
LOWCode injectioncode.eval_exec · tests/exhaustion-suite.test.ts:39
F-02 Code injectionCWE-78 · CWE-94 · CWE-95Layer L1
const fn = new Function(`${flood.code}; return typeof ${nameMatch![1]} === "function";`);
Why
evaluates text as code
Fix
remove; use a parser or a dispatch table
LOWCode injectioncode.eval_exec · tests/opaque-predicates.test.ts:9
F-02 Code injectionCWE-78 · CWE-94 · CWE-95Layer L1
return new Function("k", `return (${expr});`)(k) as boolean;
Why
evaluates text as code
Fix
remove; use a parser or a dispatch table
LOWCode injectioncode.eval_exec · tests/runtime-memory-diff-shield.test.ts:112
F-02 Code injectionCWE-78 · CWE-94 · CWE-95Layer L1
const readerFn = new Function(
Why
evaluates text as code
Fix
remove; use a parser or a dispatch table
LOWCode injectioncode.eval_exec · tests/runtime-memory-diff-shield.test.ts:160
F-02 Code injectionCWE-78 · CWE-94 · CWE-95Layer L1
const runFn = new Function(
Why
evaluates text as code
Fix
remove; use a parser or a dispatch table

Tools it exposes

33 declared in source · 0 destructive

30 read3 write0 destructive
ToolWhat it doesRisk
bla_rea_driftRun the REA constant-drift harness (scripts/verify-rea-constants.mjs). This is NOT red-team evidence. Absent sibling REA → status unverified (never a silent pass).write
bla_run_lookupRead-only correlation of a prior harden run: load bla-harden-summary.json by path or output directory, optionally asserting a run_id. Unsigned summaries only — never treats absence as success.write
shield-against-reaAudit a build artifact for REA exposure, apply BLAwrite
bla-counter-reaEvaluate REA reverse-engineering exposure and plan defensive hardening for a software artifact.read
bla-harden-projectOne-sentence Counter-REA hardening: discover every protectable build artifact under a project directory and run bla_harden (audit→defend→verify, bankrupt profile).read
bla-hardening-workflowStep-by-step release hardening checklist using bla_inspect_release, bla_audit, bla_protect, and bla_verify.read
bla-weave-decoyPlan weave-decoy hardening: read bla://directives/weave-decoy, call bla_weave_plan, then bla_harden with weave_decoys:true. Defensive only.read
bla_auditPre-flight REA exposure and vulnerability audit. Simulates REA static/semantic passes on a target file or directory (JS/TS, Electron .asar, WASM, Python .py/.pyc, Native ELF/Mach-O/PE, ZIP/APK) and estimates attacker Claude/GPT token and dollar cost.read
bla_benchmarkRuns the Counter-REA Showcase Benchmark suite and returns Before vs. After metrics.read
bla_defendOne-click Counter-REA full defense hardening pipeline. Runs release source-map exposure inspection, 16-vector REA vulnerability audit, Resource-Exhaustion + Model-Misjudgment shield injection (default: bankrupt preset), and mathematical invariant verification in a single call.read
bla_doctorChecks readiness of all BLA protection engines and agent MCP registrations.read
bla_emit_headersEmits compile-time native defense headers for C/C++ (bla_shield.h), Rust (bla_shield.rs), or Go (bla_shield.go) to defeat Ghidra/Hopper/IDA and REA native binary analyzers.read
bla_encrypt_stringsEncrypt sensitive string literals in panel.html / license.cjs with a seed-derived XOR keystream and inject a runtime decrypt stub. Reuses SeededPrng.derive(\read
bla_initCreates a starter bla.config.json in a project directory for zero-argumentread
bla_inspect_releaseRead-only local source-map and symbol exposure scan for build directories and release artifacts.read
bla_lintDEFAULT packaging-footgun gate scan for a project directory (or file). Runs every R1–R7 bankrupt refuse lint plus lane-independence and panel-derive discovery in one call — same engines as the individual bla_lint_* tools /read
bla_lint_bind_scopeRead-only lint for bla-bind-scope/v1. Refuses when public≠true, constraints empty, dates/domains invalid, hidden fields present, or bind-scope.json exists without BLA_BIND_SCOPE_CHECK in the entry. Writes nothing.read
bla_lint_integrity_coverageRead-only integrity coverage matrix lint (Direction-3). Evaluates bla-integrity-coverage/v1 JSON (path to file or directory discovering integrity-coverage.json). Refuses when full coverage / anti-single-point / hashes / connectivity fail. Hash-only — no guessing. Writes nothing.read
bla_lint_js_aes_unwrapRead-only JS AES wrap-key unwrap lint (same engine as CLIread
bla_lint_js_wasm_hydrate_dumpRead-only JS←WASM hydrate param-dump lint (same engine as CLIread
bla_lint_lanesRead-only heuristic lane-independence footgun lint (same engine as CLIread
bla_lint_panel_deriveRead-only panel-derive ABI lint (same engine as CLIread
bla_lint_plaintext_secretsRead-only lint for cleartext sensitive literals (FLAG{...}, high-entropy tokens, PEM headers, long hex/base64, derive key-name strings) in panel.html / license.cjs. Bankrupt packaging gates refuse error findings unless allow_plaintext_string. URLs are warning-class. Writes nothing.read
bla_lint_sealRead-only seal-commitment lint (same engine as CLIread
bla_lint_wasm_paramsRead-only WASM param/cipher carrier lint (same engine as CLIread
bla_lint_wasm_seal_xor_known_plaintextRead-only WASM+seal.revision XOR known-plaintext lint (same engine as CLIread
bla_lint_wasm_xor_mask_sidecarRead-only WASM XOR-mask sidecar lint (same engine as CLIread
bla_protectInjects BLAread
bla_verifyVerifies a BLA-protected artifact against REAread
directoryProject or build directory (default: .)read
target_pathPath to the build artifact or directoryread
weave-decoy-directivesChinese-friendly weave-decoy theory + executor checklist for Counter-REA agents (defensive only).read
weave-decoy-docPointer to repository theory pack ${WEAVE_DECOY_DOC_PATH}read

Safer alternatives

All security servers →

Audit history

  1. ✕BLOCK F 38 / 100

    First audit · 37cf6aa5f948

README badge

Badges show the current verdict, BLOCK included

ModelsAtlas: BLOCK FMODELSATLASBLOCK · F

A badge cannot be hidden by the maintainer while the verdict is BLOCK. Fix the findings, push, and the re-audit updates it.

Questions about bla MCP server

What is the bla MCP server?
🛡️ 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.
What tools does bla expose?
33 in total: 30 read-only, 3 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is bla safe to connect to an agent?
No — not without reading the findings first. The audit graded it F (38/100) and found 5 critical or high issues in the source. Each one is listed on this page with the file and line it is on.
What credentials does bla need?
It reads CRACKME_V6_KEY, CRACKME_V7_KEY, CRACKME_V8_KEY and CRACKME_V9_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How current is this page?
The grade is for one exact copy of the source (37cf6aa5f948), read on 2026-11-12. The repository is watched and re-audited when it changes.
I already installed bla. What should I do?
Remove it from your client's MCP configuration, then rotate any API keys and tokens that were in your environment while it ran.
Can the verdict change?
Yes. The repository is watched and re-audited when its source changes. If the findings are removed the verdict updates, and the history on this page shows it.

Sources

  • qybaihe/bla on GitHub — the source we read, its licence, stars and activityObserved · 12 Nov 2026
  • ModelsAtlas trust audit — BLOCK verdict for commit 37cf6aa5f948Observed · 12 Nov 2026

Spotted something wrong? Report a correction · Methodology

Keep going in your AI

Opens with this page as the source. We add nothing else to the prompt.