QUANTUM-CLI v4.6.3

QUANTUM-CLI v4.6.3

The operator-surface release. The runtime modules that were placeholder
dashboards are now real probes, committed, and reachable from both the
CLI and the dashboard API.

The rule that governs all of it

A module may only report a capability it actually verified.

The placeholder dashboards returned a literal hardware_connected: True
with no probe behind it — a simulated capability, which this project's
"no simulated code" rule forbids. Every module now reports what the
host actually has, and says so when it has nothing. The rule is
enforced by tests: hardware_connected may only be True when a
device was actually found.

What the modules report now

Module	Probe	Result on the reference host
cve_update	CISA KEV catalogue vs the local registry	47 of 67 registered CVEs are known-exploited
persistence_checker	Run keys, startup folders, scheduled tasks, services, systemd, cron, LD_PRELOAD, modules	731 artifacts (390 tasks, 323 services)
traffic_analysis	Interfaces, addresses, counters, capture tooling	2 interfaces, 513 MB rx / 1.9 GB tx
wallet (cryptowallet)	Bounded filesystem search for wallet artifacts	9 wallet files, 1 address, 16 keys
pivot_lateral	ARP/neighbour table + SMB, SSH, RDP, WinRM, LDAP, VNC	15 neighbours
vault_share	Volumes, Cryptomator vaults, backups, SMB shares, NFS exports	2 volumes, 1 vault, 6 shares
cookie_manager	Browser cookie stores, read from a copy, Netscape export	reads real Chromium SQLite stores
mitm_dashboard	Interception tooling, proxy env, proxy listeners, CA files	2 proxy env vars
c2_proxy	Proxy env, OS proxy config, endpoint reachability	2 env vars, egress tools found
adb_exploit	Real adb execution for every action	no device attached (reported honestly)
bluetooth / hid / sdr / imsi	BT controller, HID devices, radios, capture-capable radios + modem	1 controller, 38 HID devices, 0 radios
dark_web_browser	Tor binary, SOCKS/control ports, circuits	tor present, daemon down
nso_pegasus_ios	Device backups, MobileSync, pairing records, indicators	none found
god_mode	Real process privileges, effective user, importable modules	not admin, 5 modules
ddos	Live egress check, socket limits, NIC speed, stress tooling	egress yes, 2 tools
mobile_verification	Aggregates the mobile modules	5 verified, 5 idle
advanced_config_aggregator	Aggregates the live states	6 active, 6 idle
openrouter_config	Provider registry: endpoint, model, key present, free models, importable skills	17 free models, 19/21 skills

Reachability

CLI — 18 new operator commands, each with a --query sub-dispatch:

cve  persistence  cookiejar  mitm  traffic  c2proxy  autoconfig  godmode
vault  pegasus  otp  pivot  mobile  biometric  wallet  logclean  adbx
advanced

plus the hardware probes bluetooth, hid, sdr, imsi. Two new
arguments: --port and --mode.

API — the operator modules are exposed under /api/{crypto, biometric, otp, logclean, adb, pivot, ddos, c2proxy, autoconfig, god_mode, darkweb, vault, nso_pegasus, mitm, rat, cve, mobile, cookie, traffic, persistence, advanced} (214 routes in total).

Safety fixes

• logs/.lootkey is now immutable by construction. Deleting or
  editing it breaks every sealed .gfy in loot/. The key is refused
  by name and hidden state files are not mutable either.
• No log operation can escape logs/. The old guard used
  os.path.isabs, which does not mean the same thing on Windows —
  /etc/passwd was accepted and written outside the repository.
• State-changing API routes are POST-only. Cleaning logs, injecting
  cookies, deploying an APK and setting SMS forwarding were reachable
  with a GET, which a browser prefetch, a link preview or a crawler can
  trigger.
• No GET route reads request.json. That answers 415 for any
  client that does not send a JSON body.
• The log cleaner no longer searches the working directory by
  default. A single dashboard request once stripped a header line
  from 29 files in the source tree.
• The payload manifest records byte counts, not character counts.
  Three kernel stubs contain U+2014 — one character, three bytes — so
  the manifest disagreed with the files it described.

Other fixes

• The UiState sealed class is no longer generic, which resolves the
  Kotlin→Java CAP#1 wildcard-capture error that blocked
  compileFullReleaseJavaWithJavac.
• @JavascriptInterface methods bridge the suspend repository calls
  with runBlocking — they run on the WebView JS thread, not a
  coroutine.
• ZIP payload archives are byte-reproducible: stored rather than
  deflated (classic zlib and zlib-ng disagree byte-for-byte), a pinned
  host field, and the manifest's volatile timestamp stripped.
• build_all_kernel.py gained the --outdir the kernel-exploits
  workflow was already passing; the job had been failing on every run.
• The dirtypipe PoC defines _GNU_SOURCE and includes <stdlib.h>
  and <sys/types.h> (splice, F_GETPIPE_SZ and loff_t are GNU
  extensions).

Assets

File	Size	Notes
quantum-cli-full-4.6.3.apk	15.9 MB	V2 signed, root-capable
quantum-cli-limited-4.6.3.apk	15.9 MB	V2 signed, no-root
quantum-exploits-v4.6.3.zip	213 KB	exploit modules + operator harness
quantum-payloads-v4.6.3.zip	88 KB	64 built payload artifacts
payloads.lock.json	18 KB	sha256 manifest for every payload
PLATFORM_INDEX.json	21 KB	platform/exploit index

Both archives are byte-reproducible: every entry is stored, stamped
with a fixed epoch and a pinned host field, so rebuilding the same tree
yields identical archives. Every artifact in
quantum-payloads-v4.6.3.zip matches payloads.lock.json by size and
sha256.

They are reproducible across platforms as well: the archives this
release was built from are byte-identical to the ones a Windows
checkout produces.

Archive	sha256 (first 16)
quantum-exploits-v4.6.3.zip	8e150c825dfddd8a
quantum-payloads-v4.6.3.zip	01acf2c12b7d493a

Rebuild with:

python package_release.py --version 4.6.3

Verification

Check	Result
pytest tests/	667 passed, 10 skipped, 0 failed
ruff gate (E9,F63,F7,F82,F401,F811)	clean on the committed tree
payload manifest vs payloads.lock.json	64/64 verified
apksigner verify	V2 signature scheme on both APKs
GitHub CI	9/9 jobs green

Authorized use only

All tooling in this release assumes penetration-test, lab or
contracted-authorization contexts. Use it only against systems you own
or have written authorization to test.