v4.2.2: validation keeps reproducible findings, /pocs command, version bump

Field feedback: real, reproducible findings (missing HSTS, insecure cookie flags,
internal IP leaked in a header) were being down-rated/dropped by the adversarial
opinion-vote. The user's rule: never discard something real; another person must
be able to reproduce the same finding.

Validation:
- has_http_receipt(): a finding whose proof is a captured HTTP response (status
  line / security headers / Set-Cookie / a header the class is proven by) or a
  file:line citation is REPRODUCIBLE by definition.
- validate(): a finding unanimously rejected by the vote but carrying such a
  receipt is NO LONGER dropped — it is kept as needs-review (capped to what the
  receipt alone proves), because "the response lacks HSTS" is a fact, not a
  story. grounded_receipt() now also recognizes an in-prose HTTP receipt.
- reproducibility: a kept finding at a URL with no explicit repro steps gets a
  minimal pasteable `curl -i` so anyone can reproduce the exact finding.

Also:
- /pocs (aliases /poc /evidence /artifacts): list a run's synthesized/written
  PoC + evidence files with their paths (was: "unknown command /pocs").
- version bumped to 4.2.2 across CLI/clap/web; stale v4.2.1/v4.1.0 labels fixed.

423 tests passing.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

README.md

@@ -1,4 +1,4 @@
-<h1 align="center">🧠 NeuroSploit v4.2.1</h1>
+<h1 align="center">🧠 NeuroSploit v4.2.2</h1>
 
 <p align="center">
   <a href="https://github.com/JoasASantos/NeuroSploit/stargazers"><img src="https://img.shields.io/github/stars/JoasASantos/NeuroSploit?style=for-the-badge&logo=github&color=8b5cf6" alt="Stars"></a>
@@ -8,7 +8,7 @@
 </p>
 
 <p align="center">
-  <img src="https://img.shields.io/badge/Version-4.2.1-blue?style=flat-square">
+  <img src="https://img.shields.io/badge/Version-4.2.2-blue?style=flat-square">
   <img src="https://img.shields.io/badge/Harness-Rust%20%7C%20tokio-e6b673?style=flat-square">
   <img src="https://img.shields.io/badge/License-MIT-green?style=flat-square">
   <img src="https://img.shields.io/badge/MD%20Agents-479-red?style=flat-square">
@@ -52,7 +52,22 @@ Control TUI**.
 
 ### Highlights
 
-> **New in v4.2.1** — **SARIF 2.1.0 export**: every run now writes `report.sarif`
+> **New in v4.2.2** — **free, LLM-directed exploration**: an exploit agent's named
+> class is a starting point, not a cage — it maps what the app actually does and
+> reports any class it can prove, with **authentication / identity** (login, signup,
+> password reset, MFA, OAuth/OIDC/SAML, JWT, session) as a first-class target and
+> business-logic / multi-step flows pursued on its own judgment; agent **selection**
+> now covers the surface instead of collapsing into one family. **WAF-aware User-Agent**
+> (`/ua browser`) uses a realistic browser UA for accuracy behind a CDN while keeping
+> attribution in the `X-NeuroSploit-Scan` header. **Importable engagement configs**:
+> `/authorize <hosts…>` sets the whole scope in one line (no bug-bounty program needed),
+> `/scope-file <yaml>` imports scope **and** target/models/focus/classes from one file;
+> `/class idor,sqli,xss,ssrf` focuses a run on vuln classes. Plus **PoC/evidence files
+> synthesized from recorded evidence** even on the API-key path (empty `pocs/`·`evidence/`
+> fixed), **model-refusal detection** (a declined technique is reported as such, not a
+> parse error), and version strings read from the build so they never go stale.
+>
+> **Also in v4.2.x** — **SARIF 2.1.0 export**: every run now writes `report.sarif`
 > next to the Markdown/JSON/HTML/PDF, and `neurosploit sarif <run>` (re)emits it
 > on demand, so findings drop straight into GitHub / Azure DevOps code-scanning
 > as severity-coloured, CWE-linked alerts (also exposed over MCP). Plus stronger

TUTORIAL.md

@@ -1,4 +1,4 @@
-# NeuroSploit — Tutorial & User Guide (v4.2.1)
+# NeuroSploit — Tutorial & User Guide (v4.2.2)
 
 A complete, hands-on guide to installing, configuring and running NeuroSploit —
 the autonomous, multi-model penetration-testing harness.
@@ -100,7 +100,7 @@ Agents **degrade gracefully**: if `rustscan` is absent they use `nmap`; if neith
 ### Verify
 
 ```bash
-neurosploit --version          # neurosploit 4.2.1
+neurosploit --version          # neurosploit 4.2.2
 neurosploit agents             # {"vulns":255,...,"ai":30,...,"total":480}
 neurosploit models             # all providers & models
 ```

neurosploit-rs/Cargo.lock

@@ -940,7 +940,7 @@ dependencies = [
 
 [[package]]
 name = "neurosploit"
-version = "4.2.1"
+version = "4.2.2"
 dependencies = [
  "anyhow",
  "clap",
@@ -957,7 +957,7 @@ dependencies = [
 
 [[package]]
 name = "neurosploit-harness"
-version = "4.2.1"
+version = "4.2.2"
 dependencies = [
  "anyhow",
  "base64",

neurosploit-rs/Cargo.toml

@@ -3,7 +3,7 @@ members = ["crates/harness", "app"]
 resolver = "2"
 
 [workspace.package]
-version = "4.2.1"
+version = "4.2.2"
 edition = "2021"
 license = "MIT"
 repository = "https://github.com/JoasASantos/NeuroSploit"

neurosploit-rs/app/src/main.rs

@@ -13,8 +13,8 @@ use std::path::{Path, PathBuf};
 #[command(
     name = "neurosploit",
     version,
-    about = "NeuroSploit v4.2.1 — multi-model autonomous pentest harness",
-    long_about = "NeuroSploit v4.2.1 — a Rust multi-model harness that drives a pool of LLMs \
+    about = "NeuroSploit v4.2.2 — multi-model autonomous pentest harness",
+    long_about = "NeuroSploit v4.2.2 — a Rust multi-model harness that drives a pool of LLMs \
 (API key or local subscription: Claude/Codex/Gemini/Grok/OpenCode/Hermes) to autonomously test a target. \
 After recon it INTELLIGENTLY selects only the agents matching the discovered surface, runs \
 them in parallel, then validates every finding by cross-model voting before reporting.\n\n\

neurosploit-rs/app/src/repl.rs

@@ -148,7 +148,7 @@ struct LiveCheckpoint {
 pub(crate) const ACCEPTED: &[&str] = &[
     "/?", "/agents", "/attach", "/audit", "/auth", "/burp", "/cap", "/capability", "/chain", "/changed", "/clear", "/config",
     "/context", "/continue", "/creds", "/diff", "/exclude", "/exit", "/expand", "/feed",
-    "/finding", "/findings", "/focus", "/forget", "/full", "/go", "/goal", "/graph", "/guardrail", "/guardrails", "/help",
+    "/pocs", "/poc", "/evidence", "/artifacts", "/finding", "/findings", "/focus", "/forget", "/full", "/go", "/goal", "/graph", "/guardrail", "/guardrails", "/help",
     "/history", "/idle", "/inscope", "/instructions", "/integration", "/integrations", "/key", "/log",
     "/authorize", "/grant", "/inscope-set", "/scope-file", "/scopefile", "/import-scope", "/authorization", "/authz", "/program", "/logs", "/mcp", "/memory", "/model", "/models", "/objective", "/objectives", "/observe",
     "/observe-only", "/offline",
@@ -164,7 +164,7 @@ const COMMANDS: &[&str] = &[
     "/help", "/onboard", "/show", "/config", "/providers", "/model", "/key", "/sub", "/target",
     "/scope-file",     "/authorization", "/class",     "/repo", "/auth", "/creds", "/focus", "/objective", "/scope-out", "/attach", "/context", "/mcp", "/offline",
     "/class", "/research", "/quick", "/economy", "/eco", "/votes", "/chain", "/recon", "/tempmail", "/timeout", "/proxy", "/burp", "/ua", "/agents", "/only", "/theme", "/clear", "/run", "/stop", "/pause", "/continue", "/runs", "/results", "/report",
-    "/status", "/logs", "/diff", "/retest", "/validate", "/finding", "/expand", "/integrations",
+    "/status", "/logs", "/diff", "/retest", "/validate", "/finding", "/pocs", "/expand", "/integrations",
     "/memory", "/forget", "/graph", "/inscope", "/observe", "/guardrail", "/policy",
     "/capability", "/audit", "/quit",
 ];
@@ -1277,6 +1277,31 @@ pub async fn repl(base: &Path, auth: SessionAuth) -> anyhow::Result<()> {
                     if live_now { println!("  \x1b[2m(run still streaming in background — /logs for what happened while browsing)\x1b[0m"); }
                 }
             }
+            "/pocs" | "/poc" | "/evidence" | "/artifacts" => {
+                // List the PoC and evidence artifacts a run produced (synthesized
+                // or agent-written), with their paths so they can be retrieved.
+                let h = history.lock().unwrap();
+                let rec = if arg.trim().is_empty() { h.last() } else { pick(&h, arg) };
+                match rec {
+                    None => println!("  no run yet — /runs to list, or /pocs <id> after a run"),
+                    Some(r) if r.workdir.is_empty() => println!("  run #{} has no workdir on record", r.id),
+                    Some(r) => {
+                        let dir = std::path::Path::new(&r.workdir);
+                        let mut any = false;
+                        for (sub, label) in [("pocs", "PoC scripts"), ("evidence", "evidence")] {
+                            let p = dir.join(sub);
+                            let mut files: Vec<String> = std::fs::read_dir(&p).map(|rd| rd.filter_map(|e| e.ok()).map(|e| e.file_name().to_string_lossy().to_string()).collect()).unwrap_or_default();
+                            files.sort();
+                            if !files.is_empty() {
+                                any = true;
+                                println!("  \x1b[1m{label}\x1b[0m ({}):", p.display());
+                                for f in files { println!("    {}", p.join(&f).display()); }
+                            }
+                        }
+                        if !any { println!("  no pocs/ or evidence/ files for run #{} ({})", r.id, dir.display()); }
+                    }
+                }
+            }
             "/finding" | "/findings" => {
                 // Build the finding pool: live run if active, else a past run.
                 let pool: Vec<Finding> = match &active {

neurosploit-rs/crates/harness/src/pipeline.rs

@@ -1935,26 +1935,35 @@ async fn validate(candidates: Vec<Finding>, pool: &ModelPool, sys: &str, vote_n:
                     f.review_status = "needs-review".into();
                     f.review_reason = format!("voter rejected the narrative, mechanic retained: {}", f.review_reason.trim());
                     f.confidence = f.confidence.min(0.5);
-                } else if grounded_receipt(&f) {
-                    // Unanimously rejected, but the MECHANISM was demonstrated —
-                    // a real engagement rejected "no rate limiting on the reset
-                    // flow" because the agent claimed email flooding and only
-                    // proved that 25 requests went through unthrottled. The
-                    // claim was inflated; the measurement was real, and
-                    // discarding it hid a genuine gap from the report.
-                    //
-                    // So an over-claimed finding is capped and flagged rather
-                    // than deleted: the reader gets the fact, not the story
-                    // that was built on it.
+                } else if grounded_receipt(&f) || has_http_receipt(&f) {
+                    // Unanimously rejected by the opinion-vote, but the finding
+                    // carries a concrete, REPRODUCIBLE receipt (a captured HTTP
+                    // response, a header/cookie the class is proven by, or a
+                    // file:line citation). Another person can reproduce this
+                    // exact finding with one request — so it is NOT dropped.
+                    // The adversarial validator is tuned to reject low-impact and
+                    // "theoretical" issues, but "the response lacks HSTS" or "the
+                    // cookie has no Secure flag" is a FACT, not a story. We cap
+                    // the severity to what the receipt alone proves and flag it
+                    // for human review, never delete it.
                     let cap = "Low";
                     let was = f.severity.clone();
-                    f.severity = cap.to_string();
+                    if sev_rank(&f.severity) < sev_rank(cap) { f.severity = cap.to_string(); }
                     f.review_status = "needs-review".into();
-                    f.review_reason = format!(
-                        "impact not demonstrated — capped from {was} to {cap}. Validator: {}",
-                        f.review_reason.trim()
-                    );
-                    f.confidence = f.confidence.min(0.5);
+                    f.review_reason = if was == f.severity {
+                        format!("kept — reproducible receipt present; impact not independently demonstrated. Validator: {}", f.review_reason.trim())
+                    } else {
+                        format!("kept & capped from {was} to {} — reproducible receipt present, impact not demonstrated. Validator: {}", f.severity, f.review_reason.trim())
+                    };
+                    f.confidence = f.confidence.max(0.3).min(0.6);
+                }
+                // Reproducibility: a kept finding that reaches a URL endpoint but
+                // carries no explicit repro steps gets a minimal, pasteable one,
+                // so another person can reproduce the exact finding.
+                if (f.validated || f.review_status == "needs-review")
+                    && f.repro_steps.is_empty()
+                    && (f.endpoint.starts_with("http://") || f.endpoint.starts_with("https://")) {
+                    f.repro_steps = vec![format!("curl -i -s '{}'   # inspect the response (status + headers + body) that proves this finding", f.endpoint)];
                 }
                 let label = if f.validated { "CONFIRMED" } else if f.review_status == "needs-review" { "needs-review" } else { "rejected" };
                 let _ = txc.send(format!("vote {} → {} ({})", f.title, label, f.votes)).await;
@@ -2154,9 +2163,39 @@ fn grounded_receipt(f: &Finding) -> bool {
     if f.review_reason.contains("receipt_missing") || f.votes.contains("receipt_missing") {
         return false;
     }
+    if has_http_receipt(f) {
+        return true;
+    }
     crate::grounding::ground(f, "", crate::grounding::GroundMode::Either).ok
 }
 
+/// True when the finding's evidence carries a concrete, reproducible HTTP
+/// receipt — a captured response (status line / headers / Set-Cookie) or a
+/// `file:line` code citation. This is the test for "another person can
+/// reproduce this exact finding": if the proof is the response itself (a missing
+/// security header, an insecure cookie flag, an internal IP leaked in a header,
+/// a status code), it is reproducible with a single request and must never be
+/// discarded by an opinion-based vote — demoted to needs-review at worst.
+fn has_http_receipt(f: &Finding) -> bool {
+    if f.evidence_data.is_some() { return true; }
+    let hay = format!("{}\n{}", f.evidence, f.payload).to_lowercase();
+    // A captured HTTP response or the headers/fields these deterministic classes
+    // are proven by.
+    const SIGNS: &[&str] = &[
+        "http/1.1", "http/2", "http/1.0", "status: ", "status code", "status=",
+        "set-cookie", "strict-transport-security", "x-frame-options",
+        "content-security-policy", "access-control-allow-origin", "location:",
+        "server:", "www-authenticate", "< http", "=> http", "response:", "200 ok",
+        "301 ", "302 ", "401 ", "403 ", "404 ", "500 ", "curl ",
+    ];
+    if SIGNS.iter().any(|s| hay.contains(s)) { return true; }
+    // A white-box file:line citation is also a reproducible receipt.
+    f.endpoint.contains(':') && f.endpoint.chars().any(|c| c.is_ascii_digit())
+        && (f.endpoint.contains(".rs") || f.endpoint.contains(".py") || f.endpoint.contains(".js")
+            || f.endpoint.contains(".php") || f.endpoint.contains(".java") || f.endpoint.contains(".go")
+            || f.endpoint.contains(".ts") || f.endpoint.contains(".rb"))
+}
+
 /// Adversarial refutation pass: every confirmed **High/Critical** finding is
 /// re-examined by a skeptical panel that tries to prove it's a false positive.
 /// A finding that fails to withstand a majority of skeptics is dropped. Lower
@@ -4440,6 +4479,34 @@ mod extraction_tests {
         assert!(!looks_like_refusal("No vulnerabilities were found in the tested endpoints."));
     }
 
+    #[test]
+    fn a_response_backed_finding_is_a_reproducible_receipt_and_never_dropped() {
+        // Missing HSTS: the proof is the response headers — reproducible with one
+        // request, must never be discarded by an opinion-vote.
+        let hsts = Finding {
+            title: "No Strict-Transport-Security header".into(),
+            severity: "Low".into(), cwe: "CWE-319".into(),
+            endpoint: "https://scapi.rockstargames.com/".into(),
+            evidence: "HTTP/2 200\nserver: cloudflare\n(no strict-transport-security header present)".into(),
+            ..Default::default()
+        };
+        assert!(has_http_receipt(&hsts), "a captured response IS a receipt");
+        // Insecure cookie flags — Set-Cookie in evidence.
+        let cookie = Finding {
+            title: "Cookie without Secure/HttpOnly".into(), severity: "Low".into(), cwe: "CWE-614".into(),
+            endpoint: "https://scapi.rockstargames.com/".into(),
+            evidence: "Set-Cookie: bal=1; path=/   (no Secure, no HttpOnly, no SameSite)".into(),
+            ..Default::default()
+        };
+        assert!(has_http_receipt(&cookie));
+        // Pure prose with no receipt is NOT a reproducible receipt.
+        let vague = Finding { title: "Maybe vulnerable".into(), evidence: "the app seems insecure".into(), ..Default::default() };
+        assert!(!has_http_receipt(&vague));
+        // A white-box file:line citation IS a receipt.
+        let wb = Finding { title: "SQLi".into(), endpoint: "src/db.py:42".into(), evidence: "query = f\"...{id}\"".into(), ..Default::default() };
+        assert!(has_http_receipt(&wb));
+    }
+
     /// The bare-array happy path for agent selection.
     #[test]
     fn a_string_array_parses_plain() {

web/public/index.html

@@ -3,7 +3,7 @@
 <head>
 <meta charset="utf-8" />
 <meta name="viewport" content="width=device-width, initial-scale=1" />
-<title>NeuroSploit v4.2.1 — Console</title>
+<title>NeuroSploit v4.2.2 — Console</title>
 <link rel="icon" href="data:image/svg+xml,<svg xmlns=%22http://www.w3.org/2000/svg%22 viewBox=%220 0 100 100%22><text y=%22.9em%22 font-size=%2290%22>🧠</text></svg>">
 <link rel="stylesheet" href="/vendor/xterm.css" />
 <link rel="stylesheet" href="/style.css" />
@@ -33,7 +33,7 @@
     <div class="sb-groups" id="sbGroups"><!-- populated by app.js --></div>
 
     <div class="sb-bottom">
-      <span class="sb-version" id="sbVersion">v4.2.1</span>
+      <span class="sb-version" id="sbVersion">v4.2.2</span>
       <div class="sb-bottom-actions">
         <button class="icon-btn" id="btnOpenAuth" title="Auth &amp; API keys">🔑</button>
         <button class="icon-btn" id="btnOpenRepl" title="Open terminal (Ctrl+`)">❭_</button>

web/server.js

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 'use strict';
 /**
- * NeuroSploit v4.2.1 — web console backend.
+ * NeuroSploit v4.2.2 — web console backend.
  *
  * Zero-dependency Node HTTP server that:
  *  - serves the static SPA in ./public
@@ -1487,7 +1487,7 @@ const server = http.createServer(async (req, res) => {
     }
 
     if (req.method === 'GET' && p === '/api/meta') {
-      return sendJson(res, 200, { version: '4.2.1', binary: BIN, root: ROOT });
+      return sendJson(res, 200, { version: "4.2.2", binary: BIN, root: ROOT });
     }
 
     // ---- providers / API keys (in-memory only, never persisted) ----
@@ -1521,7 +1521,7 @@ const server = http.createServer(async (req, res) => {
 loadPersistedJobs();
 
 server.listen(PORT, () => {
-  console.log(`NeuroSploit v4.2.1 web console → http://localhost:${PORT}`);
+  console.log(`NeuroSploit v4.2.2 web console → http://localhost:${PORT}`);
   console.log(`  binary : ${BIN || '(not found — build neurosploit-rs first)'}`);
   console.log(`  agents : ${AGENTS_DIR}`);
   console.log(`  runs   : ${RUNS_DIR}`);