eb4e13efea
v3.5.1: live findings + /finding + Ctrl+O/expand + 3-way /stop (soft validate) + report URL + structured Typst + IIS/CMS/CVE agents
...
REPL interactivity & findings:
- Live findings registered during a run: /results shows them accumulating;
/finding opens a selection menu with FULL details (PoC, command, evidence,
CVSS, OWASP/CWE, remediation). Past runs too.
- /expand (and Ctrl+O) dump the last full, untruncated commands.
- Findings colored by severity in the feed (not all-yellow); confirmed vote = green.
Stop & report:
- CRITICAL: /stop no longer kills validation. New SOFT stop (pool.soft) halts
launching new agents but lets in-flight + VALIDATION finish — so confirmed
findings are kept. /stop now asks 3 ways: [1] validate then report,
[2] report raw (no validation), [3] discard.
- Report file:// URL printed on completion/stop.
Report:
- Typst report restructured: executive summary, a Vulnerability Summary TABLE
(#, vuln, severity, CVSS, OWASP/CWE), and per-finding sections with criticality,
CVSS, OWASP/CWE, description/impact, PoC, evidence, remediation. owasp passed through.
Agents: +14 app-stack/CVE (IIS tilde/WebDAV/ViewState/debug/handler-bypass,
CMS fingerprint + WordPress/Joomla/Drupal/default-admin, app-server consoles,
exposed VCS, known-CVE & outdated-component exploitation) → 343 total.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com >
2026-06-24 23:21:43 -03:00
639c2209f7
v3.5.1: attack-chain agents (12) + per-project .neurosploit/ persistence & resume
...
Chaining:
- agents_md/chains/ (12 multi-stage exploitation playbooks): SQLi→RCE→LPE,
SSRF→AWS-creds, SSRF→RCE, upload→RCE, upload→LFI→RCE→LPE, XSS→ATO, IDOR→ATO,
SSTI→RCE→cloud, default-creds→domain, deserialization→RCE, exposed-git→RCE,
subdomain-takeover→trusted-abuse. Each stage proven by a tool receipt before
advancing; reports chains_from edges.
- Loaded as a `chains` category (→ 329 agents). chain_round now injects the chain
recipes as a menu so the LLM applies proven multi-stage paths.
Persistence (no DB — structured state):
- Per-project `<cwd>/.neurosploit/` holding session.json (config), runs.json
(history), history.txt (readline). REPL resumes target/repo/auth/focus/models
on reopen; saves on /run and /quit.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com >
2026-06-24 22:30:22 -03:00
f8d70ce9c5
v3.5.1: infra/host engagements — IP + SSH/Windows-AD creds + Linux/Win/AD agents + REPL context bar
...
Infra:
- creds.yaml gains `ssh:` (host/port/user/password/key) and `windows:`/`ad:`
(host/user/password/domain/ntlm-hash) blocks; multi-block YAML parser.
host_instruction() tells agents how to authenticate to the host.
- 14 infra agents (agents_md/infra/): port/service scan, SMB enum, Linux privesc/
sudo/cron/SSH, Windows privesc/SMB-signing/WinRM, AD kerberoast/asreproast/ACL/
DCSync/default-creds. Loader gains `infra` category → 317 agents total.
- run_host pipeline + `neurosploit host <ip> --creds creds.yaml` (and Mode::Host
in run_mode/TUI): host recon (nmap/netexec) → infra agent selection → test →
validate → chain → report, with host tooling doctrine + supplied creds.
REPL:
- Context/status bar above the prompt: "model auth · cwd · mode▸target"
(e.g. claude-opus-4-8 sub · /opt/projeto · black-box▸app.acme.com).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com >
2026-06-24 22:17:14 -03:00
5d83e8848e
v3.4.1: harness intelligence — router, ReAct, dedup, token-trim, configurable MCP, +54 code agents, credits
...
- Task-based model ROUTER (recon/select prefer a fast model; exploit prefers primary; validate uses a different model than the finder)
- ReAct doctrine injected into exploit prompts (Thought→Action→Observation, token-efficient)
- Dedup: unique agents per run + findings deduped by CWE/endpoint/title (highest confidence kept)
- Token economy: recon blob capped for selector + per-agent context
- Configurable MCP: merge user mcp.servers.json into the pipeline's .mcp.json
- +54 white-box/code-analysis agents (NoSQLi, LDAP/XPath, JWT-none, Java/.NET/PHP/Go/Node/Python
specifics, SSTI, ReDoS, deserialization, etc.) → 303 agents total (78 code)
- Credits: Joas A Santos & Red Team Leaders (CLI banner, interactive header, HTML+Typst report)
- README: GitHub stars/forks badges, 60-second quick start, full API config steps, intuitive layout
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com >
2026-06-24 19:49:01 -03:00
3ca3f269ee
v3.4.x: intelligent agent selection, whitebox, recon/code agents, Gemini, artifacts, RL, XBOW GUI
...
Harness intelligence:
- After recon, the model SELECTS which specialist agents match the target
(select_agents) — runs the relevant subset, not blindly top-N
- RL reward store (rl.rs): per-agent weights persist to data/rl_state_rs.json,
reward validated findings (severity-weighted), decay idle, bias next run
- Run artifacts persisted as JSON + MD (recon, exploitation transcript,
findings, html report) under runs/<target>-<ts>/ for reuse by other AIs
Whitebox mode:
- run_whitebox: walks a repo, builds bounded source context, runs code agents,
validates by adversarial vote. CLI `whitebox <path>` + web "White-box" mode
Agents: +12 recon (subdomain/tech/js/api/secrets/dns/content/param/waf/cloud/
graphql/osint) and +24 code SAST reviewers (sqli/cmdi/path/ssrf/xss/deser/
secrets/crypto/authz/idor/xxe/redirect/ssti/race/eval/csrf/random/logging/
upload/mass-assign/jwt/cors). Loader gains recon/ + code/ categories → 249 total
Models: +Google Gemini provider (API + gemini CLI subscription); installed_cli_
backends now detects gemini; chat_cli handles gemini/codex/grok + optional
Playwright MCP (.mcp.json) on the subscription path with autonomy flags
GUI: full XBOW-style redesign — sidebar (Operate/Library), topbar status, mode
segment (black-box/white-box), model panel, live console, severity cards,
agent browser with category filters, models view; responsive + aligned
Verified: cargo build --release clean; CLI agents/whitebox; LIVE subscription
run shows model selecting 23→4 agents, RL update, artifacts written; GUI +
white-box toggle in Playwright.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com >
2026-06-23 11:39:56 -03:00
55af0d4634
NeuroSploit v3.3.0 — Autonomous MD-Agent Engine
...
Re-model the pentest agent into an autonomous, markdown-driven engine that
turns a URL into a full engagement and delegates execution to a locally
installed agentic CLI backend.
Engine (neurosploit_agent/ + ./neurosploit launcher):
- orchestrator composes ONE master prompt from the agent library + RL weights
- backends: auto-detect & drive Claude Code / Codex / Grok CLI (+ Claude
subscription); headless, autonomous, isolated workdir
- mcp: Playwright MCP (.mcp.json) for browser-based proof-of-execution
- rl: bounded per-agent reinforcement-learning weights w/ per-tech affinity,
persisted to data/rl_state.json
- models: latest registry incl. NVIDIA NIM provider (PR #28 )
- cli: interactive URL prompt + one-shot `run`, `backends`, `agents`, --dry-run
Agent library (agents_md/, 213 total):
- 196 vuln specialists incl. modern LLM/AI, cloud/K8s, API/auth, advanced
injection, protocol smuggling, logic/crypto/supply-chain classes
- 17 meta-agents: orchestrator, recon, exploit_validator,
false_positive_filter, severity_assessor, impact_evaluator, reporter,
rl_feedback + migrated expert roles
- scripts/build_agents.py data-driven builder; REGISTRY.md index
Docs: rewritten README.md, v3.3.0 RELEASE.md, .env.example (NVIDIA NIM, xAI,
engine vars).
Retire legacy Python orchestration (neurosploit.py + agent classes) to legacy/.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com >
2026-06-14 20:57:38 -03:00
CyberSecurityUP