Files
NeuroSploit/web/server.js
T
CyberSecurityUPandClaude Sonnet 5 4fbe608a7a feat(web): drive run/whitebox/greybox exploitation through a real REPL session
Root cause of "can't send prompts while a run streams": /api/exploit
spawned a plain `neurosploit run ...` subprocess, and that CLI path
(run_mode() in main.rs) never reads stdin - it only waits on the task or
Ctrl-C. The ONLY thing in the harness that keeps accepting input while an
engagement streams is the interactive REPL's background-run loop. So:

- New startJobViaRepl(): for mode run/whitebox/greybox, spawns a bare
  `neurosploit` REPL session and scripts it via stdin (/target or /repo,
  /model, /sub, /mcp, /votes, /chain, /recon, /focus, /objective,
  /scope-out, /creds, /only <agents> or /only clear, then /run) instead
  of building CLI args. Same underlying pipeline, same tagged output
  lines, so all existing parsing (findings/phase/progress/runId) works
  unchanged. host/aitest/skills modes stay on the old one-shot
  startJob() - they need onboarding's scope picker, an interactive
  arrow-key menu that silently skips itself over a piped stdin, so they
  can't be scripted this way.
- New POST /api/exploit/:id/input writes a line to the session's stdin -
  natural language, /status, /continue, anything the REPL accepts - and
  the live run view grows a "send prompt" box (in the Activity log tab)
  for it, shown only when the job reports interactive: true.
- Stop, for an interactive job, now sends the REPL's own graceful
  '/stop\n1\n' (validate what's found, then report) instead of SIGINT -
  the REPL's own input loop has no signal handler, so SIGINT there would
  just kill the process outright and skip the report step. Non-
  interactive jobs still get SIGINT (run_mode() does catch that).
- 'done' can no longer be process-exit only: an interactive session stays
  open after the engagement finishes (for /report, /continue, another
  /run), so ingestLine() now also flags done from the same "phase
  complete" content signal it already used for the phase field.

Verified end-to-end: started an interactive job, confirmed
`interactive: true` and a captured runId, sent /status and /agents mid-
and post-run over the new /input endpoint (both accepted, session stayed
alive and responsive after completion), and confirmed a non-interactive
run is unaffected.

Also: the missing "Activity log" tab a screenshot showed for a "running"
engagement was the sidebar's detail-view fallback (2 tabs, no log) for a
run whose Job object no longer exists in server memory - it happens when
the Node process gets restarted while a spawned neurosploit child is
still alive underneath it (an orphan from testing across many redeploys
this session, not a code bug); the live view itself always had the tab.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0129WdYHccPsH27k5GGuwijd
2026-08-23 16:04:55 -03:00

971 lines
42 KiB
JavaScript

#!/usr/bin/env node
'use strict';
/**
* NeuroSploit v4.0.0 — web console backend.
*
* Zero-dependency Node HTTP server that:
* - serves the static SPA in ./public
* - reads agents_md/ to build the "lead board" (agent/category picker)
* - reads runs/ to build run history + structured findings
* - spawns the compiled `neurosploit` CLI binary for exploitation runs and
* streams its stdout (parsed into structured events) over SSE
* - spawns the interactive `neurosploit` REPL (no subcommand) as a child
* process and pipes stdin/stdout so the browser gets a REAL REPL
* connected to the CLI harness — not a reimplementation.
*/
const http = require('node:http');
const fs = require('node:fs');
const fsp = fs.promises;
const os = require('node:os');
const path = require('node:path');
const { spawn } = require('node:child_process');
const crypto = require('node:crypto');
const { EventEmitter } = require('node:events');
// ---------------------------------------------------------------------------
// Paths
// ---------------------------------------------------------------------------
const WEB_DIR = __dirname;
const ROOT = path.resolve(WEB_DIR, '..'); // repo root — holds agents_md/, runs/
const AGENTS_DIR = path.join(ROOT, 'agents_md');
const RUNS_DIR = path.join(ROOT, 'runs');
const NAMES_FILE = path.join(ROOT, '.neurosploit', 'web-engagement-names.json');
// Engagement names are set by the operator in the wizard before launch (not
// something the CLI/harness knows about) — persisted here as runId -> name so
// the sidebar/run history can label a run by its engagement name across
// restarts, not just by target/run-id.
const engagementNames = new Map();
try {
const raw = JSON.parse(fs.readFileSync(NAMES_FILE, 'utf8'));
for (const [k, v] of Object.entries(raw)) engagementNames.set(k, v);
} catch { /* no file yet — fine */ }
function saveEngagementName(runId, name) {
if (!runId || !name) return;
engagementNames.set(runId, name);
fsp.mkdir(path.dirname(NAMES_FILE), { recursive: true })
.then(() => fsp.writeFile(NAMES_FILE, JSON.stringify(Object.fromEntries(engagementNames), null, 2)))
.catch(() => {});
}
const PUBLIC_DIR = path.join(WEB_DIR, 'public');
function findBinary() {
const candidates = [
path.join(ROOT, 'neurosploit-rs', 'target', 'release', 'neurosploit'),
path.join(ROOT, 'neurosploit-rs', 'target', 'debug', 'neurosploit'),
];
for (const c of candidates) {
if (fs.existsSync(c)) return c;
}
return null;
}
const BIN = findBinary();
const PORT = Number(process.env.NEUROSPLOIT_WEB_PORT || process.env.PORT || 4173);
// ---------------------------------------------------------------------------
// Providers — mirrors crates/harness/src/models.rs `providers()`. Kept as a
// literal table (not parsed from CLI output) so `kind` ("cli" = usable via a
// locally-installed agentic CLI subscription login, "api" = key-only) and
// `envKey` (the environment variable the harness reads for that provider)
// are available without shelling out. Keep this in sync when models.rs adds
// a provider.
// ---------------------------------------------------------------------------
const PROVIDERS = [
{ key: 'anthropic', label: 'Anthropic Claude', kind: 'cli', envKey: 'ANTHROPIC_API_KEY',
models: ['claude-opus-5', 'claude-sonnet-5', 'claude-opus-4-8', 'claude-sonnet-4-6', 'claude-haiku-4-5'] },
{ key: 'openai', label: 'OpenAI (ChatGPT)', kind: 'cli', envKey: 'OPENAI_API_KEY',
models: ['gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.6-luna', 'gpt-5.5', 'gpt-5.4', 'gpt-5.4-mini', 'gpt-5.3-codex', 'gpt-5.2', 'gpt-5.1', 'gpt-5.1-codex', 'o4'] },
{ key: 'xai', label: 'xAI Grok', kind: 'cli', envKey: 'XAI_API_KEY',
models: ['grok-4.5', 'grok-4', 'grok-4-fast'] },
{ key: 'gemini', label: 'Google Gemini', kind: 'cli', envKey: 'GEMINI_API_KEY',
models: ['gemini-3-pro', 'gemini-2.5-pro', 'gemini-2.5-flash'] },
{ key: 'opencode', label: 'OpenCode Zen', kind: 'cli', envKey: 'OPENCODE_API_KEY',
models: ['claude-opus-5', 'claude-sonnet-5', 'gpt-5.6-sol', 'gpt-5.5', 'gemini-3-pro', 'grok-4.5', 'deepseek-v4-pro', 'qwen3.7-max', 'kimi-k3'] },
{ key: 'nous', label: 'Nous Research (Hermes)', kind: 'cli', envKey: 'NOUS_API_KEY',
models: ['Hermes-4-405B', 'Hermes-4-70B', 'DeepHermes-3-Mistral-24B-Preview'] },
{ key: 'nvidia_nim', label: 'NVIDIA NIM', kind: 'api', envKey: 'NVIDIA_NIM_API_KEY',
models: ['nvidia/llama-3.3-nemotron-super-49b-v1', 'deepseek-ai/deepseek-r1', 'qwen/qwen2.5-coder-32b-instruct'] },
{ key: 'deepseek', label: 'DeepSeek', kind: 'api', envKey: 'DEEPSEEK_API_KEY',
models: ['deepseek-reasoner', 'deepseek-chat'] },
{ key: 'mistral', label: 'Mistral', kind: 'api', envKey: 'MISTRAL_API_KEY',
models: ['mistral-large-latest', 'codestral-latest'] },
{ key: 'qwen', label: 'Qwen (DashScope)', kind: 'api', envKey: 'DASHSCOPE_API_KEY',
models: ['qwen-max', 'qwen2.5-coder-32b-instruct', 'qwq-plus'] },
{ key: 'groq', label: 'Groq', kind: 'api', envKey: 'GROQ_API_KEY',
models: ['llama-3.3-70b-versatile', 'qwen-2.5-coder-32b'] },
{ key: 'together', label: 'Together AI', kind: 'api', envKey: 'TOGETHER_API_KEY',
models: ['Qwen/Qwen2.5-Coder-32B-Instruct', 'deepseek-ai/DeepSeek-R1', 'meta-llama/Llama-3.3-70B-Instruct-Turbo'] },
{ key: 'moonshot', label: 'Moonshot AI (Kimi)', kind: 'api', envKey: 'MOONSHOT_API_KEY',
models: ['kimi-k3', 'kimi-k2', 'moonshot-v1-128k', 'moonshot-v1-32k'] },
{ key: 'litellm', label: 'LiteLLM (proxy)', kind: 'api', envKey: 'LITELLM_API_KEY',
models: ['gpt-4o', 'claude-3-7-sonnet', 'gemini/gemini-2.5-pro'] },
{ key: 'openrouter', label: 'OpenRouter', kind: 'api', envKey: 'OPENROUTER_API_KEY',
models: ['anthropic/claude-opus-4-8', 'qwen/qwen-2.5-coder-32b-instruct', 'deepseek/deepseek-r1', 'meta-llama/llama-3.3-70b-instruct'] },
{ key: 'azure', label: 'Azure OpenAI', kind: 'api', envKey: 'AZURE_OPENAI_API_KEY',
models: ['gpt-4o', 'gpt-4o-mini', 'gpt-5.1', 'o4-mini'] },
{ key: 'ollama', label: 'Ollama (local)', kind: 'api', envKey: 'OLLAMA_API_KEY',
models: ['qwen2.5-coder:32b', 'qwq:32b', 'deepseek-r1:32b', 'llama3.3:70b'] },
{ key: 'llamacpp', label: 'llama.cpp (local)', kind: 'api', envKey: 'LLAMACPP_API_KEY',
models: ['qwen2.5-coder-32b-instruct', 'dolphin-2.9-llama3-70b', 'deepseek-r1-distill-qwen-32b', 'llama-3.3-70b-instruct'] },
];
// In-memory only — never written to disk. Cleared on server restart.
const apiKeys = new Map(); // provider key -> secret
function envOverrides() {
const env = {};
for (const [key, secret] of apiKeys) {
const p = PROVIDERS.find((x) => x.key === key);
if (p && secret) env[p.envKey] = secret;
}
return env;
}
/// Build a minimal creds.yaml-compatible file (see neurosploit-rs/creds.example.yaml)
/// from a raw auth header and/or named roles, so the CLI's --creds flag can be
/// used to carry web-entered auth material without a real file on disk.
function buildCredsYaml({ auth, roles }) {
const lines = ['# generated by the NeuroSploit web console — ephemeral, not committed'];
if (auth) lines.push(`header: ${JSON.stringify(auth)}`);
for (const r of roles || []) {
if (!r?.name || !r?.header) continue;
const safe = String(r.name).replace(/[^a-zA-Z0-9_-]/g, '_');
lines.push(`${safe}:`, ` header: ${JSON.stringify(r.header)}`);
}
return lines.join('\n') + '\n';
}
async function materializeCreds(body, jobId) {
if (body.creds) return body.creds; // explicit file path on disk wins
if (!body.auth && !(body.roles || []).length) return undefined;
const dir = path.join(os.tmpdir(), 'neurosploit-web');
await fsp.mkdir(dir, { recursive: true });
const file = path.join(dir, `${jobId}.creds.yaml`);
await fsp.writeFile(file, buildCredsYaml(body));
return file;
}
// ---------------------------------------------------------------------------
// Agent library — read agents_md/{vulns,ai,infra,code,chains,recon,meta}/*.md
// and classify each into a lead category the UI can group + toggle.
// ---------------------------------------------------------------------------
const KIND_DIRS = {
vulns: 'vuln',
ai: 'ai',
infra: 'infra',
code: 'code',
chains: 'chain',
recon: 'recon',
meta: 'meta',
};
// Ordered classifier: first matching rule wins. Mirrors the vocabulary used
// throughout agents_md/ so every agent lands in a sensible bucket for the
// lead board (screenshot-style category groups).
const CATEGORY_RULES = [
[/^(llm_|mcp_|n8n_|redteam_|skill_|prompt_injection|ml_model_inversion|vector_db_injection)/, 'LLM Application'],
[/^(xss_|dom_xss|blind_xss|mutation_xss|dom_clobbering|postmessage_vulnerability)/, 'Cross-Site Scripting'],
[/^(jwt_|oauth_|oidc_|saml_|session_fixation|mfa_bypass|two_factor|twofa_|captcha_bypass|brute_force|default_credentials|weak_password|weak_jwt_secret|login_sqli_bypass|timing_side_channel_auth|timing_attack|refresh_token_abuse|auth_bypass|password_reset_poisoning)/, 'Auth & Session'],
[/^(idor|bola|bfla|access_control_bypass|privilege_escalation|forced_browsing|authenticated_surface_exploit|exposed_admin_panel|spa_hidden_admin|mass_assignment|register_privilege_mass_assign|api_excessive_data|excessive_data_exposure|account_takeover_chain)/, 'Broken Access Control'],
[/^(business_logic|spa_business_logic|coupon_logic_abuse|price_manipulation|workflow_step_skip|race_condition|idempotency_key_abuse|account_registration_and_forms)/, 'Business Logic'],
[/^(sqli_|nosql_injection|ldap_injection|xpath_injection|xslt_injection|ssti|command_injection|log_injection|crlf_injection|header_injection|email_injection|smtp_injection|soap_injection|orm_injection|expression_language_injection|graphql_injection|css_injection|html_injection|csv_injection|formula_injection_excel|dangling_markup_injection|client_side_template_injection|server_side_prototype_pollution|prototype_pollution|xxe|path_traversal|^lfi$|^rfi$|zip_slip|log4shell_jndi|pickle_deserialization|insecure_deserialization|yaml_deserialization|type_juggling)/, 'Injection'],
[/^(ssrf|gcp_metadata_ssrf|azure_imds_exposure|aws_imds_v2_bypass|host_header_injection|http_smuggling|http_desync|http2_request_smuggling|h2c_smuggling|reverse_proxy_path_confusion|websocket_)/, 'SSRF & Network'],
[/^(graphql_|api_rate_limiting|rest_api_versioning|api_key_exposure|exposed_api_docs|spa_api_discovery|param_miner|parameter_pollution|grpc_reflection_exposure|api_bola)/, 'API & GraphQL'],
[/^(aws_|azure_|gcp_|s3_bucket|gcs_bucket_misconfig|k8s_|docker_socket_exposure|container_escape|cloud_|terraform_state_exposure|helm_secret_exposure|ecr_public_exposure|serverless_|ci_cd_secret_leak|ad_)/, 'Cloud & Infra'],
[/^(clickjacking|tabnabbing|cors_misconfig|insecure_cookie_flags|security_headers|subdomain_takeover|open_redirect|second_order_redirect|oauth_open_redirect_chain|csrf)/, 'Client-Side'],
[/^(weak_encryption|weak_hashing|weak_random|padding_oracle|ecb_pattern_leak|ssl_issues|cleartext_transmission)/, 'Cryptography'],
[/^(rate_limit|graphql_dos|regex_dos|range_header_dos|web_cache_poisoning_dos|llm_model_dos)/, 'Rate Limiting & DoS'],
[/^(cache_poisoning|cdn_cache_key_poisoning|web_cache_deception|insecure_cdn|byte_range_cache|edge_side_includes)/, 'Cache & CDN'],
[/^(cve_|eol_|version_disclosure|wordpress_audit|joomla_audit|drupal_audit|cms_|outdated_|dependency_confusion|typosquatting_package|vulnerable_dependency|git_exposed_repo|git_svn_exposure_app|source_code_disclosure|backup_file_exposure|env_file_exposure|debug_mode|aspnet_|appserver_exposure|misconfig_|iis_|information_disclosure|sensitive_data_exposure|directory_listing)/, 'Recon & Fingerprint'],
[/^linux_/, 'Linux Host'],
[/^windows_/, 'Windows Host'],
];
function classify(name, kind) {
if (name.startsWith('custom_')) return 'Custom Leads'; // generated via /api/leads/generate
if (kind === 'chain') return 'Attack Chains';
if (kind === 'recon') return 'Recon';
if (kind === 'code') return 'Code Review';
if (kind === 'meta') return 'Meta & Reporting';
if (kind === 'ai') return 'LLM Application';
for (const [re, cat] of CATEGORY_RULES) {
if (re.test(name)) return cat;
}
return 'Other';
}
function extractTitle(text, fallback) {
const m = text.match(/^#\s+(.+?)\s*$/m);
return m ? m[1].trim() : fallback;
}
function extractCwe(text) {
const m = text.match(/CWE-\d+/);
return m ? m[0] : '';
}
let agentCache = null;
let agentCacheAt = 0;
async function loadAgents() {
const now = Date.now();
if (agentCache && now - agentCacheAt < 5000) return agentCache;
const agents = [];
for (const [dir, kind] of Object.entries(KIND_DIRS)) {
const full = path.join(AGENTS_DIR, dir);
let entries = [];
try {
entries = await fsp.readdir(full);
} catch {
continue;
}
for (const file of entries) {
if (!file.endsWith('.md')) continue;
const name = file.slice(0, -3);
const text = await fsp.readFile(path.join(full, file), 'utf8').catch(() => '');
agents.push({
id: name,
name,
title: extractTitle(text, name),
cwe: extractCwe(text),
kind,
category: classify(name, kind),
});
}
}
agents.sort((a, b) => a.name.localeCompare(b.name));
const byCategory = new Map();
for (const a of agents) {
if (!byCategory.has(a.category)) byCategory.set(a.category, []);
byCategory.get(a.category).push(a);
}
// Selectable leads only (exclude meta/orchestration from the pentest board —
// they're internal doctrine agents, not testable "leads").
const LEAD_ORDER = [
'Custom Leads',
'Business Logic', 'Broken Access Control', 'Injection', 'Cross-Site Scripting',
'LLM Application', 'Auth & Session', 'SSRF & Network', 'API & GraphQL',
'Cloud & Infra', 'Client-Side', 'Cryptography', 'Rate Limiting & DoS',
'Cache & CDN', 'Recon & Fingerprint', 'Linux Host', 'Windows Host',
'Attack Chains', 'Code Review', 'Recon', 'Other',
];
const categories = LEAD_ORDER
.filter((c) => byCategory.has(c) && c !== 'Meta & Reporting')
.map((c) => ({ category: c, agents: byCategory.get(c) }));
agentCache = { total: agents.length, agents, categories };
agentCacheAt = now;
return agentCache;
}
// ---------------------------------------------------------------------------
// Custom leads — "+ Custom lead" generates a REAL specialist-agent markdown
// file (same format agents_md/vulns/*.md uses) via the `claude` CLI on the
// operator's Anthropic subscription, so a custom lead is an actual pinnable
// agent, not just free text folded into --focus. Mirrors the exact one-shot
// invocation harness::models::cli_login_status() uses for the same CLI.
// ---------------------------------------------------------------------------
const GEN_MODEL = 'claude-opus-4-8'; // matches Session::default() in app/src/repl.rs
const GEN_TIMEOUT_MS = 90_000;
function slugify(s) {
return (s || '').toLowerCase().replace(/[^a-z0-9]+/g, '_').replace(/^_+|_+$/g, '').slice(0, 40) || 'lead';
}
function buildSkillGenPrompt(description) {
return `Write ONE new security-testing specialist-agent file for this custom lead, in EXACTLY this markdown shape and nothing else — no code fences around the whole thing, no preamble, no explanation, just the file content starting at the first line:
# <Short Title> Agent
## User Prompt
You are testing **{target}** for <the specific vulnerability class this lead targets>.
**Recon Context:**
{recon_json}
**METHODOLOGY:**
### 1. <step name>
- <concrete technique>
### 2. <step name>
- <concrete technique>
(as many numbered steps as the vuln class actually needs — terse, technical, no filler)
### Report
\`\`\`
FINDING:
- Title: ...
- Severity: ...
- CWE: CWE-<pick the single most fitting CWE number>
- Endpoint: [URL]
- Evidence: ...
- Impact: ...
- Remediation: ...
\`\`\`
## System Prompt
<one paragraph: the agent's persona plus the ONE calibration rule that stops it from reporting a finding without real proof>
The operator's custom lead request, verbatim: "${description}"
Match the doctrine style of NeuroSploit's other agents_md/vulns/*.md files: terse, technical, no marketing language, one CWE, a real report template.
Do not use any tools (no file writes, no bash, no search) — this is a pure text-completion task. Respond with ONLY the markdown file content above, nothing before it and nothing after it.`;
}
function generateCustomLead(description) {
return new Promise((resolve, reject) => {
if (!binaryOnPath('claude')) {
return reject(new Error("claude CLI not found on PATH — install Claude Code and run `claude` to log in first"));
}
// No --dangerously-skip-permissions here: this is a pure text-completion
// call (no bash/file tools needed), and granting tool access made claude
// try to write the file itself and narrate doing so instead of just
// returning text — see buildSkillGenPrompt()'s explicit "no tools" line.
const child = spawn('claude', ['-p', '--model', GEN_MODEL, '--output-format', 'text'], { env: process.env });
let out = '', err = '';
const timer = setTimeout(() => { child.kill('SIGKILL'); reject(new Error('generation timed out')); }, GEN_TIMEOUT_MS);
child.stdout.on('data', (c) => { out += c; });
child.stderr.on('data', (c) => { err += c; });
child.on('error', (e) => { clearTimeout(timer); reject(new Error(`claude CLI failed to start: ${e.message}`)); });
child.on('close', (code) => {
clearTimeout(timer);
if (!out.trim()) return reject(new Error(err.trim() || `claude exited ${code} with no output — is it logged in? run \`claude\` once to check.`));
resolve(out);
});
child.stdin.write(buildSkillGenPrompt(description));
child.stdin.end();
});
}
function binaryOnPath(bin) {
const dirs = (process.env.PATH || '').split(path.delimiter);
return dirs.some((d) => { try { return fs.existsSync(path.join(d, bin)); } catch { return false; } });
}
// ---------------------------------------------------------------------------
// Runs — read runs/<id>/{meta,status,findings}.json
// ---------------------------------------------------------------------------
async function readJsonSafe(p, fallback) {
try {
return JSON.parse(await fsp.readFile(p, 'utf8'));
} catch {
return fallback;
}
}
async function listRuns() {
let ids = [];
try {
ids = (await fsp.readdir(RUNS_DIR)).filter((d) => d.startsWith('ns-'));
} catch {
return [];
}
const runs = await Promise.all(ids.map(async (id) => {
const dir = path.join(RUNS_DIR, id);
const [meta, status, findings] = await Promise.all([
readJsonSafe(path.join(dir, 'meta.json'), {}),
readJsonSafe(path.join(dir, 'status.json'), {}),
readJsonSafe(path.join(dir, 'findings.json'), []),
]);
const tsMatch = id.match(/^ns-(\d+)-/);
const ts = tsMatch ? Number(tsMatch[1]) : 0;
const sevCount = {};
for (const f of findings) sevCount[f.severity] = (sevCount[f.severity] || 0) + 1;
return {
id,
ts,
name: engagementNames.get(id) || '',
target: status.target || meta.target || id.replace(/^ns-\d+-/, ''),
state: status.state || 'unknown',
findings: findings.length,
severities: sevCount,
hasReport: fs.existsSync(path.join(dir, 'report.html')) || fs.existsSync(path.join(dir, 'report.pdf')),
};
}));
runs.sort((a, b) => b.ts - a.ts);
return runs;
}
async function runDetail(id) {
const dir = safeRunDir(id);
if (!dir) return null;
const [meta, status, findings] = await Promise.all([
readJsonSafe(path.join(dir, 'meta.json'), {}),
readJsonSafe(path.join(dir, 'status.json'), {}),
readJsonSafe(path.join(dir, 'findings.json'), []),
]);
const assets = ['report.html', 'report.pdf', 'report.md', 'recon.md', 'exploitation.md']
.filter((f) => fs.existsSync(path.join(dir, f)));
const pocs = await fsp.readdir(path.join(dir, 'pocs')).catch(() => []);
return { id, name: engagementNames.get(id) || '', meta, status, findings, assets, pocs };
}
function safeRunDir(id) {
if (!/^[a-zA-Z0-9_.-]+$/.test(id)) return null;
const dir = path.join(RUNS_DIR, id);
if (!dir.startsWith(RUNS_DIR)) return null;
return dir;
}
// ---------------------------------------------------------------------------
// Exploitation jobs — spawn `neurosploit <mode> <target> --only ... -v`
// and parse its stdout into structured live state (mirrors app/src/repl.rs
// RunLive::ingest so the web UI gets the same phases/findings the TUI does).
// ---------------------------------------------------------------------------
const jobs = new Map(); // id -> Job
class Job extends EventEmitter {
constructor(id, cmd, args, target, name) {
super();
this.id = id;
this.cmd = cmd;
this.args = args;
this.target = target || '';
this.name = name || '';
this.pinnedAgents = [];
this.runId = null; // ns-<ts>-<target> workdir basename, once known
this.phase = 'starting';
this.findings = [];
this.feed = [];
this.agents = 0;
this.agentsDone = 0;
this.done = false;
this.exitCode = null;
this.reportUrl = null;
this.startedAt = Date.now();
this.child = null;
}
push(evt) {
this.feed.push(evt);
if (this.feed.length > 2000) this.feed.shift();
this.emit('event', evt);
}
snapshot() {
return {
id: this.id,
target: this.target,
name: this.name,
pinnedAgents: this.pinnedAgents,
interactive: !!this.repl,
runId: this.runId,
phase: this.phase,
findings: this.findings,
agents: this.agents,
agentsDone: this.agentsDone,
done: this.done,
exitCode: this.exitCode,
reportUrl: this.reportUrl,
startedAt: this.startedAt,
};
}
}
const ANSI_RE = /\x1b\[[0-9;]*[A-Za-z]/g;
function stripAnsi(s) { return s.replace(ANSI_RE, ''); }
function ingestLine(job, rawLine) {
const line = stripAnsi(rawLine);
const low = line.toLowerCase();
job.push({ type: 'log', line });
if (low.includes('token/quota exhausted') || low.includes('run is paused')) job.phase = 'paused (quota)';
else if (low.includes('authentication failed') || low.includes('circuit breaker')) job.phase = 'paused (auth)';
else if (low.startsWith('recon') || low.startsWith('ai-recon') || low.includes('recon round') || low.startsWith('probe:')) job.phase = 'recon';
else if (low.includes('selected') && low.includes('agent')) {
job.phase = 'planning';
const n = line.split(/\s+/).map(Number).find((x) => Number.isFinite(x));
if (n) job.agents = n;
} else if (low.startsWith('exploit') || low.startsWith('test ') || low.includes('launching agent')) job.phase = 'exploiting';
else if (low.startsWith('vote') || low.includes('validating')) job.phase = 'validating';
else if (low.startsWith('chain')) job.phase = 'chaining';
else if (low.includes('phase complete') || low.includes('validated finding(s)')) {
job.phase = 'complete';
// A REPL-backed job's child process doesn't exit when the engagement
// finishes (the session stays open for /report, /continue, another
// /run, ...) — so 'done' has to come from content, not process exit.
if (job.repl && !job.done) { job.done = true; job.push({ type: 'done', exitCode: 0 }); }
}
if (/candidate\(s\)/.test(low) && /^(exploit |test |analyze |review )/.test(low)) job.agentsDone += 1;
const fj = line.match(/^finding_json:\s*(.+)$/);
if (fj) {
try {
const finding = JSON.parse(fj[1]);
job.findings.push(finding);
job.push({ type: 'finding', finding });
} catch { /* ignore malformed line */ }
}
const rep = line.match(/report:\s*(file:\/\/\S+)/);
if (rep) job.reportUrl = rep[1];
const rid = line.match(/run id\s*:\s*(\S+)/);
if (rid) { job.runId = rid[1]; saveEngagementName(job.runId, job.name); }
}
function buildArgs(body) {
const mode = body.mode || 'run';
const args = [mode];
if (mode === 'run' || mode === 'host' || mode === 'aitest') {
args.push(body.target);
} else if (mode === 'whitebox' || mode === 'skills') {
args.push(body.repo || body.target);
} else if (mode === 'greybox') {
args.push(body.repo);
args.push('--url', body.target);
}
for (const m of body.models || []) args.push('--model', m);
if (body.votes) args.push('--vote-n', String(body.votes));
if (body.chainDepth !== undefined) args.push('--chain-depth', String(body.chainDepth));
if (body.recon) args.push('--recon', String(body.recon));
if (body.maxAgents) args.push('--max-agents', String(body.maxAgents));
if (body.offline) args.push('--offline');
if (body.subscription) args.push('--subscription');
if (body.mcp) args.push('--mcp');
if (body.creds) args.push('--creds', body.creds);
if (body.focus) args.push('--focus', body.focus);
if (body.objective) args.push('--objective', body.objective);
if (body.outOfScope) args.push('--out-of-scope', body.outOfScope);
for (const a of body.agents || []) args.push('--only', a);
args.push('--verbose');
return args;
}
async function startJob(body) {
if (!BIN) throw new Error('neurosploit binary not found — run `cargo build --release` in neurosploit-rs/');
const id = crypto.randomUUID();
const credsPath = await materializeCreds(body, id);
const args = buildArgs({ ...body, creds: credsPath });
const job = new Job(id, BIN, args, body.repo || body.target || '', body.name || '');
job.pinnedAgents = body.agents || [];
jobs.set(id, job);
const child = spawn(BIN, args, { cwd: ROOT, env: { ...process.env, ...envOverrides() } });
job.child = child;
let buf = '';
const onData = (chunk) => {
buf += chunk.toString('utf8');
let idx;
while ((idx = buf.indexOf('\n')) !== -1) {
const line = buf.slice(0, idx);
buf = buf.slice(idx + 1);
if (line.length) ingestLine(job, line);
}
};
child.stdout.on('data', onData);
child.stderr.on('data', onData);
child.on('close', (code) => {
if (buf.trim()) ingestLine(job, buf);
job.done = true;
job.exitCode = code;
job.phase = job.phase === 'paused (quota)' || job.phase === 'paused (auth)' ? job.phase : 'complete';
job.push({ type: 'done', exitCode: code });
});
child.on('error', (err) => {
job.done = true;
job.push({ type: 'log', line: `[web] failed to start neurosploit: ${err.message}` });
job.push({ type: 'done', exitCode: -1 });
});
return job;
}
/// Turn the wizard's config into the REPL commands that produce the same
/// engagement (`/target`/`/repo` → `/model` → toggles → `/only` → `/run`).
/// `/only` is what makes this equivalent to the CLI's `--only` — REPL had no
/// such command before this feature (added to app/src/repl.rs alongside it).
function buildReplScript(body) {
const lines = [];
if (body.mode === 'whitebox') lines.push(`/repo ${body.repo || body.target}`);
else {
if (body.target) lines.push(`/target ${body.target}`);
if (body.mode === 'greybox' && body.repo) lines.push(`/repo ${body.repo}`);
}
if ((body.models || []).length) lines.push(`/model ${body.models.join(',')}`);
lines.push(`/sub ${body.subscription ? 'on' : 'off'}`);
lines.push(`/mcp ${body.mcp ? 'on' : 'off'}`);
if (body.votes) lines.push(`/votes ${body.votes}`);
if (body.chainDepth !== undefined) lines.push(`/chain ${body.chainDepth}`);
if (body.recon) lines.push(`/recon ${body.recon}`);
if (body.focus) lines.push(`/focus ${body.focus}`);
if (body.objective) lines.push(`/objective ${body.objective}`);
if (body.outOfScope) lines.push(`/scope-out ${body.outOfScope}`);
if (body.creds) lines.push(`/creds ${body.creds}`);
lines.push((body.agents || []).length ? `/only ${body.agents.join(',')}` : '/only clear');
lines.push('/run');
return lines;
}
/// Same job abstraction as startJob(), but driven through a REAL interactive
/// REPL session instead of a one-shot `neurosploit run ...` subprocess — the
/// engagement streams identically (same underlying pipeline, same tagged
/// lines), but the session KEEPS reading stdin while it runs, so the web UI
/// can send more input mid-run (natural language, /status, /stop, /continue)
/// via POST /api/exploit/:id/input. Only run/whitebox/greybox support this —
/// host/aitest/skills need onboarding's scope picker, which is an interactive
/// arrow-key menu that skips itself entirely over a piped stdin.
async function startJobViaRepl(body) {
if (!BIN) throw new Error('neurosploit binary not found — run `cargo build --release` in neurosploit-rs/');
const id = crypto.randomUUID();
const credsPath = await materializeCreds(body, id);
const script = buildReplScript({ ...body, creds: credsPath });
const job = new Job(id, BIN, [], body.repo || body.target || '', body.name || '');
job.pinnedAgents = body.agents || [];
job.repl = true;
jobs.set(id, job);
const child = spawn(BIN, [], { cwd: ROOT, env: { ...process.env, ...envOverrides() } });
job.child = child;
let buf = '';
const onData = (chunk) => {
buf += chunk.toString('utf8');
let idx;
while ((idx = buf.indexOf('\n')) !== -1) {
const line = buf.slice(0, idx);
buf = buf.slice(idx + 1);
if (line.length) ingestLine(job, line);
}
};
child.stdout.on('data', onData);
child.stderr.on('data', onData);
// The REPL process itself exiting (e.g. after /quit) is ALSO a valid done
// signal, in addition to the content-based one in ingestLine().
child.on('close', (code) => {
if (buf.trim()) ingestLine(job, buf);
if (!job.done) { job.done = true; job.push({ type: 'done', exitCode: code }); }
job.exitCode = code;
});
child.on('error', (err) => {
job.done = true;
job.push({ type: 'log', line: `[web] failed to start neurosploit: ${err.message}` });
job.push({ type: 'done', exitCode: -1 });
});
for (const line of script) child.stdin.write(line + '\n');
return job;
}
// ---------------------------------------------------------------------------
// REPL sessions — spawn `neurosploit` with no subcommand (Reader::Plain kicks
// in over a piped stdin) and forward stdin/stdout verbatim: a real REPL.
// ---------------------------------------------------------------------------
const replSessions = new Map();
class ReplSession extends EventEmitter {
constructor(id, child) {
super();
this.id = id;
this.child = child;
this.done = false;
this.buffer = [];
}
push(chunk) {
this.buffer.push(chunk);
if (this.buffer.length > 5000) this.buffer.shift();
this.emit('data', chunk);
}
}
function startRepl() {
if (!BIN) throw new Error('neurosploit binary not found — run `cargo build --release` in neurosploit-rs/');
const id = crypto.randomUUID();
const child = spawn(BIN, [], { cwd: ROOT, env: { ...process.env, ...envOverrides() } });
const session = new ReplSession(id, child);
replSessions.set(id, session);
const onData = (chunk) => session.push(stripAnsi(chunk.toString('utf8')));
child.stdout.on('data', onData);
child.stderr.on('data', onData);
child.on('close', (code) => {
session.done = true;
session.push(`\n[repl session ended, exit code ${code}]\n`);
session.emit('close');
});
child.on('error', (err) => {
session.done = true;
session.push(`\n[failed to start neurosploit: ${err.message}]\n`);
session.emit('close');
});
return session;
}
// ---------------------------------------------------------------------------
// Tiny HTTP plumbing (no framework)
// ---------------------------------------------------------------------------
function sendJson(res, code, obj) {
const body = JSON.stringify(obj);
res.writeHead(code, {
'Content-Type': 'application/json; charset=utf-8',
'Content-Length': Buffer.byteLength(body),
'Cache-Control': 'no-store',
});
res.end(body);
}
function readBody(req) {
return new Promise((resolve, reject) => {
let data = '';
req.on('data', (c) => { data += c; if (data.length > 5_000_000) req.destroy(); });
req.on('end', () => {
if (!data) return resolve({});
try { resolve(JSON.parse(data)); } catch (e) { reject(e); }
});
req.on('error', reject);
});
}
function sseInit(res) {
res.writeHead(200, {
'Content-Type': 'text/event-stream; charset=utf-8',
'Cache-Control': 'no-cache',
Connection: 'keep-alive',
'X-Accel-Buffering': 'no',
});
res.write(':ok\n\n');
}
function sseSend(res, event, data) {
res.write(`event: ${event}\ndata: ${JSON.stringify(data)}\n\n`);
}
const MIME = {
'.html': 'text/html; charset=utf-8',
'.js': 'text/javascript; charset=utf-8',
'.css': 'text/css; charset=utf-8',
'.json': 'application/json; charset=utf-8',
'.svg': 'image/svg+xml',
'.png': 'image/png',
'.pdf': 'application/pdf',
'.md': 'text/plain; charset=utf-8',
};
async function serveStatic(req, res, urlPath) {
let rel = urlPath === '/' ? '/index.html' : urlPath;
const full = path.join(PUBLIC_DIR, rel);
if (!full.startsWith(PUBLIC_DIR)) { res.writeHead(403); res.end(); return; }
try {
const data = await fsp.readFile(full);
const ext = path.extname(full);
res.writeHead(200, { 'Content-Type': MIME[ext] || 'application/octet-stream' });
res.end(data);
} catch {
res.writeHead(404);
res.end('not found');
}
}
async function serveRunAsset(req, res, id, rest) {
const dir = safeRunDir(id);
if (!dir) { res.writeHead(400); res.end(); return; }
const full = path.join(dir, rest);
if (!full.startsWith(dir)) { res.writeHead(403); res.end(); return; }
try {
const data = await fsp.readFile(full);
const ext = path.extname(full);
res.writeHead(200, { 'Content-Type': MIME[ext] || 'application/octet-stream' });
res.end(data);
} catch {
res.writeHead(404);
res.end('not found');
}
}
const server = http.createServer(async (req, res) => {
const u = new URL(req.url, 'http://localhost');
const p = u.pathname;
try {
// ---- static ----
if (req.method === 'GET' && !p.startsWith('/api/')) {
return serveStatic(req, res, p);
}
// ---- agents / lead board ----
if (req.method === 'GET' && p === '/api/agents') {
return sendJson(res, 200, await loadAgents());
}
if (req.method === 'POST' && p === '/api/leads/generate') {
const body = await readBody(req);
const description = (body.description || '').trim();
if (!description) return sendJson(res, 400, { error: 'description is required' });
let raw;
try {
raw = await generateCustomLead(description);
} catch (e) {
return sendJson(res, 502, { error: e.message });
}
// Defensive: discard any wrapper text before the first '# ' heading —
// a model with tool access sometimes narrates ("I'll write the file
// now...") before the actual content despite being told not to.
const titleIdx = raw.search(/^#\s+/m);
if (titleIdx === -1) {
return sendJson(res, 502, { error: 'generation did not return a well-formed agent file', raw: raw.slice(0, 800) });
}
const clean = raw.slice(titleIdx).trim();
const titleMatch = clean.match(/^#\s+(.+?)\s*$/m);
if (!titleMatch || !/##\s*User Prompt/i.test(clean) || !/##\s*System Prompt/i.test(clean)) {
return sendJson(res, 502, { error: 'generation did not return a well-formed agent file', raw: raw.slice(0, 800) });
}
const slug = `custom_${slugify(titleMatch[1])}`;
await fsp.writeFile(path.join(AGENTS_DIR, 'vulns', `${slug}.md`), clean + '\n');
agentCache = null; // force a fresh read so the new lead shows up immediately
const { agents } = await loadAgents();
const created = agents.find((a) => a.id === slug);
return sendJson(res, 200, { agent: created, raw: clean });
}
// ---- runs ----
if (req.method === 'GET' && p === '/api/runs') {
return sendJson(res, 200, await listRuns());
}
let m = p.match(/^\/api\/runs\/([^/]+)$/);
if (req.method === 'GET' && m) {
const detail = await runDetail(decodeURIComponent(m[1]));
if (!detail) return sendJson(res, 404, { error: 'run not found' });
return sendJson(res, 200, detail);
}
m = p.match(/^\/api\/runs\/([^/]+)\/asset\/(.+)$/);
if (req.method === 'GET' && m) {
return serveRunAsset(req, res, decodeURIComponent(m[1]), decodeURIComponent(m[2]));
}
// ---- exploitation jobs ----
if (req.method === 'GET' && p === '/api/exploit') {
return sendJson(res, 200, [...jobs.values()].map((j) => j.snapshot()));
}
if (req.method === 'POST' && p === '/api/exploit') {
const body = await readBody(req);
// run/whitebox/greybox go through a real REPL session so the operator
// can keep sending it input while it streams; host/aitest/skills need
// the onboarding scope picker (an interactive menu that only works on
// a real TTY), so they stay on the plain one-shot CLI subprocess.
const replCapable = ['run', 'whitebox', 'greybox'].includes(body.mode || 'run');
const job = replCapable ? await startJobViaRepl(body) : await startJob(body);
return sendJson(res, 200, { id: job.id, interactive: !!job.repl });
}
m = p.match(/^\/api\/exploit\/([^/]+)$/);
if (req.method === 'GET' && m) {
const job = jobs.get(m[1]);
if (!job) return sendJson(res, 404, { error: 'job not found' });
return sendJson(res, 200, job.snapshot());
}
m = p.match(/^\/api\/exploit\/([^/]+)\/stop$/);
if (req.method === 'POST' && m) {
const job = jobs.get(m[1]);
if (!job) return sendJson(res, 404, { error: 'job not found' });
if (job.repl && job.child?.stdin?.writable) {
// The REPL's own graceful stop: /stop then choose "1" — validate
// what's found so far, then report. Plain SIGINT doesn't map to
// anything here (no signal handler in the REPL's own input loop).
job.child.stdin.write('/stop\n1\n');
} else {
job.child?.kill('SIGINT');
}
return sendJson(res, 200, { ok: true });
}
m = p.match(/^\/api\/exploit\/([^/]+)\/input$/);
if (req.method === 'POST' && m) {
const job = jobs.get(m[1]);
if (!job) return sendJson(res, 404, { error: 'job not found' });
if (!job.repl || !job.child?.stdin?.writable) {
return sendJson(res, 409, { error: 'this job is not an interactive session (host/aitest/skills engagements run non-interactively)' });
}
const body = await readBody(req);
job.child.stdin.write(String(body.line ?? '') + '\n');
return sendJson(res, 200, { ok: true });
}
m = p.match(/^\/api\/exploit\/([^/]+)\/events$/);
if (req.method === 'GET' && m) {
const job = jobs.get(m[1]);
if (!job) { res.writeHead(404); return res.end(); }
sseInit(res);
// replay what already happened
for (const evt of job.feed) sseSend(res, evt.type, evt);
sseSend(res, 'snapshot', job.snapshot());
if (job.done) { sseSend(res, 'done', job.snapshot()); res.end(); return; }
const onEvt = (evt) => sseSend(res, evt.type, evt);
job.on('event', onEvt);
const ping = setInterval(() => res.write(':ping\n\n'), 20000);
req.on('close', () => { job.off('event', onEvt); clearInterval(ping); });
return;
}
// ---- REPL (real CLI harness session) ----
if (req.method === 'POST' && p === '/api/repl') {
const session = startRepl();
return sendJson(res, 200, { id: session.id });
}
m = p.match(/^\/api\/repl\/([^/]+)\/input$/);
if (req.method === 'POST' && m) {
const session = replSessions.get(m[1]);
if (!session) return sendJson(res, 404, { error: 'session not found' });
const body = await readBody(req);
session.child.stdin.write(String(body.line ?? '') + '\n');
return sendJson(res, 200, { ok: true });
}
m = p.match(/^\/api\/repl\/([^/]+)\/stop$/);
if (req.method === 'POST' && m) {
const session = replSessions.get(m[1]);
if (!session) return sendJson(res, 404, { error: 'session not found' });
session.child.kill('SIGTERM');
return sendJson(res, 200, { ok: true });
}
m = p.match(/^\/api\/repl\/([^/]+)\/events$/);
if (req.method === 'GET' && m) {
const session = replSessions.get(m[1]);
if (!session) { res.writeHead(404); return res.end(); }
sseInit(res);
for (const chunk of session.buffer) sseSend(res, 'data', { chunk });
if (session.done) { sseSend(res, 'close', {}); res.end(); return; }
const onData = (chunk) => sseSend(res, 'data', { chunk });
const onClose = () => { sseSend(res, 'close', {}); res.end(); };
session.on('data', onData);
session.on('close', onClose);
const ping = setInterval(() => res.write(':ping\n\n'), 20000);
req.on('close', () => { session.off('data', onData); session.off('close', onClose); clearInterval(ping); });
return;
}
if (req.method === 'GET' && p === '/api/meta') {
return sendJson(res, 200, { version: '4.0.0', binary: BIN, root: ROOT });
}
// ---- providers / API keys (in-memory only, never persisted) ----
if (req.method === 'GET' && p === '/api/providers') {
return sendJson(res, 200, PROVIDERS.map(({ key, label, kind, models }) => ({ key, label, kind, models })));
}
if (req.method === 'GET' && p === '/api/keys') {
return sendJson(res, 200, PROVIDERS.map((pr) => ({ provider: pr.key, set: apiKeys.has(pr.key) && !!apiKeys.get(pr.key) })));
}
if (req.method === 'POST' && p === '/api/keys') {
const body = await readBody(req);
if (!body.provider || !PROVIDERS.some((pr) => pr.key === body.provider)) {
return sendJson(res, 400, { error: 'unknown provider' });
}
if (body.key) apiKeys.set(body.provider, body.key);
else apiKeys.delete(body.provider);
return sendJson(res, 200, { ok: true });
}
m = p.match(/^\/api\/keys\/([^/]+)$/);
if (req.method === 'DELETE' && m) {
apiKeys.delete(decodeURIComponent(m[1]));
return sendJson(res, 200, { ok: true });
}
sendJson(res, 404, { error: 'not found' });
} catch (err) {
sendJson(res, 500, { error: err.message });
}
});
server.listen(PORT, () => {
console.log(`NeuroSploit v4.0.0 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}`);
});