Security Enthusiast | OSCP & Google Cybersecurity Professional Certtified | CVE Holder
Poland
Joined May 2019
- Tweets95
- Following105
- Followers62
- Likes681
Pinned Tweet
Wystąpię na #CyberThreat2026 w Londynie z prezentacją „Inside Telegram’s Russian Payment Gate Network”. Pokażę, jak mapować rosyjskie bramki płatnicze na Telegramie i wykorzystywać OSINT do detekcji oszustw.
sans.org/cyber-security-trai…
🤖 Made with AI
Node2Shell PoC is now public.
CVE-2026-111497 — memory corruption in brotli 1.2.0 affecting Node.js 22/24/26, leading to RCE.
No execSync. No eval. No fake webhook.
The brotli decoder itself can reach system().
PoC:
github.com/syrex1013/CVE-202…
#Node2Shell #NodeJS #CVE #InfoSec
Built an MCP server for InstaBox, cloud Linux machines for AI agents.
Claude, Codex, Antigravity and omp can now create a box, run code, edit files, snapshot it and publish a port. 30 tools + slash commands, read-only mode, MIT.
github.com/syrex1013/instabo…
🐛 PHP json_encode() UAF — new variant (unreported)
stdClass + JsonSerializable callback → zend_hash_do_resize() fires mid-FOREACH → old arData freed → dangling read
Affects PHP 8.4.x / 8.5.x (confirmed 8.5.5)
Reported to [email protected]
PoC: github.com/syrex1013/php-jso…
Published my PHP2Shell PoC.
CVE-2026-109618 / CVE-2026-109619 are critical PHP vulnerabilities associated with unauthenticated remote code execution.
PoC / research:
github.com/syrex1013/php2she…
For security research and authorized testing only.
If you use AI coding agents, check this out.
AgentRouter is handing out $200 in free credits to try:
• GLM-5.3
• DeepSeek V4 Flash
• Claude Opus 4.8
It also works with Claude Code, OMP, Codex, Pi, DeepSeek Harness and more.
→ Sign up: agentrouter.org/register?aff…
Use an active GitHub account to qualify for the bonus.
#AI #CodingAgents #ClaudeCode #DeepSeek #OpenCode #codex
ALT AGENTROUTER dash
If you use AI coding agents, check this out.
AgentRouter is handing out $200 in free credits to try:
• GLM-5.3
• DeepSeek V4 Flash
• Claude Opus 4.8
It also works with Claude Code, OMP, Codex, Pi, DeepSeek Harness and more.
→ Sign up: agentrouter.org/register?aff…
Use an active GitHub account to qualify for the bonus.
#AI #CodingAgents #ClaudeCode #DeepSeek #OpenCode #codex
ALT AGENTROUTER dash
If you use AI coding agents, check this out.
AgentRouter is handing out $200 in free credits to try:
• GLM-5.3
• DeepSeek V4 Flash
• Claude Opus 4.8
It also works with Claude Code, OMP, Codex, Pi, DeepSeek Harness and more.
→ Sign up: agentrouter.org/register?aff…
Use an active GitHub account to qualify for the bonus.
#AI #CodingAgents #ClaudeCode #DeepSeek #OpenCode #codex
I run FutureToken. I published a small, provider-neutral repo showing the OpenAI Python SDK and curl migration: change the base_url, select a model from the live catalog, and keep the client shape.
github.com/syrex1013/futuret…
Source docs: futuretoken.threatsoft.org/d…
If your app already uses the OpenAI SDK, switching gateways can be a config change: use FutureToken's OpenAI-compatible quickstart, keep your client code, and choose an exact model ID from the live catalog. Docs: futuretoken.threatsoft.org/d… #AI #LLM
A practical API rollout checklist: start with a small key, set spend and rate limits, test streaming/tool calls, then inspect token usage. FutureToken documents the API surface, limits, and usage workflow here: futuretoken.threatsoft.org/d… #AI #API
Separate vendor keys and balances make multi-model projects harder to operate. FutureToken exposes supported models through one OpenAI-compatible /v1 endpoint, one key, and one credit balance. Capabilities and model IDs: futuretoken.threatsoft.org/m… #OpenAI #DevTools
Current FutureToken low-cost routes from the live catalog: DeepSeek V4.1 Flash $0.09/$0.36, GPT-5.6 Luna $0.10/$0.60, and Qwen3.8 Flash $0.1425/$0.446495 per 1M input/output tokens. Same OpenAI-compatible /v1 format, one key/balance: futuretoken.threatsoft.org/m…
Catalog check: FutureToken /models currently lists 9 models. DeepSeek V4.1 Flash: USD 0.09 input / 0.36 output per 1M; GPT-5.6 Luna: USD 0.10 / 0.60; Qwen3.8 Flash: USD 0.1425 / 0.446495. The catalog is the source of truth; it currently has no Gemini row. futuretoken.threatsoft.org/m…
Coding with agents? Set base_url=futuretoken.threatsoft.org/v… once. Then call Claude Sonnet 5, DeepSeek Flash, GPT-5.6 Luna — all under one balance. No per-model key juggling. Streaming supported. Pricing on /models.
Set your base_url to futuretoken.threatsoft.org/v…, drop in your key, and call any of 9 models in one SDK. Streaming, tool calls, spend controls. No per-model juggling. Live catalog on /models with prices updated by the minute.
FutureToken vs LiteLLM vs Bifrost: same problem, different approach. OpenAI-compatible endpoint? Done. Multi-provider? Yes. One balance/API key with live per-token pricing on /models and failed requests not billed? That's the difference. Coding agents: swap base_url.
FutureToken: one OpenAI-compatible API, 9 models (Claude Opus/Sonnet 5, GPT-5.6, DeepSeek V4.1 Flash, GLM-5.3, Qwen3.8). One key, one balance, one bill. Coding agents: change base_url, keep the SDK. Live prices on /models.
Technical snapshot of FutureToken (OpenAI-compatible gateway + live model table): gist.github.com/syrex1013/6d…
Source of truth for prices: futuretoken.threatsoft.org/m…
Coding agents (Claude Code, Cursor, custom OpenAI clients): point base_url at FutureToken and keep your SDK.
Live gateway prices (USD / 1M tokens):
• DeepSeek V4.1 Flash $0.09 / $0.36
• GPT-5.6 Luna $0.10 / $0.60
• claude-sonnet-5 $1.00 / $5.00
• Claude Opus 5 $3.00 / $15.00
One key, one balance. Catalog: futuretoken.threatsoft.org/m…
Catalog + live prices: futuretoken.threatsoft.org/m…
Quickstart (OpenAI SDK base_url): futuretoken.threatsoft.org/d…