# Pwnry · 2342 Pwnry GmbH

> Full machine-readable overview for AI search, LLM retrieval, and agents.

## Summary
24x7 Breach Simulation, externe Pentests mit KI-Inferenz ausschließlich in Nürnberg und automatisierte Sicherheitsprüfungen für KI-Agenten-Skills. Recurring external penetration testing with AI inference exclusively in Nuremberg and automated AI agent skill security checks.

## Description
DE: Pwnry ermittelt die öffentlich sichtbare Angriffsfläche einer Firma und bietet begrenzte automatisierte Prüfungen nach ausdrücklicher Freigabe. EN: Pwnry discovers a company's external attack surface and offers bounded automated checks with explicit authorization. Full penetration tests require a separate agreement.

## Audience
Unternehmen mit öffentlichen Webanwendungen, APIs und IT-Diensten; Empfehlungspartner. Companies with public applications, APIs and services; referral partners.

## Website
https://pwnry.ai/

## Important Pages
- [KI-Skill Scanner (DE)](https://pwnry.ai/de/skill-checks): Claude-Code-, Codex- und Gemini-Skill-Pakete vor der Installation automatisiert prüfen. Private Berichte mit Belegen und Grenzen.
- [AI Skill Scanner (EN)](https://pwnry.ai/en/skill-checks): Automated inspection of Claude Code, Codex and Gemini skill packages before installation. Private evidence reports with coverage limits.
- [Agent Skills sicherer installieren (DE)](https://pwnry.ai/de/ratgeber/agent-skills-sicherheit): SKILL.md, Skripte, Zugangsdaten-Verweise und Berechtigungen: Checkliste mit Primärquellen.
- [Agent skill security checklist (EN)](https://pwnry.ai/en/guides/agent-skill-security): SKILL.md, scripts, credential references and permissions: a practical pre-installation checklist with primary sources.
- [Franz Bettag: Forschung und Nachweise (DE)](https://pwnry.ai/de/nachweise): Historisches HackerOne-VDP-Ranking Deutschland Q1/2026, Originalbeleg und öffentliche Research-Credits.
- [Franz Bettag: research and evidence (EN)](https://pwnry.ai/en/evidence): Historical Germany HackerOne VDP ranking for Q1 2026, original evidence and public research credits.
- [Research-Fallstudie (DE)](https://pwnry.ai/de/fallstudien/bug-bounty): Untersuchung, praktische Gegenprüfung und belegte Reports in Franz Bettags eigener Research-Plattform.
- [Research case study (EN)](https://pwnry.ai/en/case-studies/bug-bounty): Investigation, practical verification and substantiated reports in Franz Bettag’s own research platform.
- [24x7 Breach Simulation (DE)](https://pwnry.ai/de/breach-simulation): BAS, externe Angriffswege, Prüfrhythmen, Umfang und Kosten.
- [24x7 Breach Simulation (EN)](https://pwnry.ai/en/breach-simulation): BAS, external attack paths, testing cadence, scope and pricing.
- [Kostenloser Firmencheck (DE)](https://pwnry.ai/de/kostenloser-sicherheitscheck): Firmen-E-Mail bestätigen; DNS-Nachweis für zusätzliche, große oder gemeinsam genutzte Domains.
- [Free company check (EN)](https://pwnry.ai/en/free-security-check): Verify company email; DNS proof is required for additional domains, shared providers and large scopes.
- [Externe Pentests (DE)](https://pwnry.ai/de/pentests): Scope, Bewertung, Behebung und Nachprüfung.
- [External penetration tests (EN)](https://pwnry.ai/en/pentests): Scope, assessment, remediation and retesting.
- [Kosten und Intervalle (DE)](https://pwnry.ai/de/preise): Unverbindliche Kalkulation nach Prüfumfang und Rhythmus.
- [Pricing and frequency (EN)](https://pwnry.ai/en/pricing): Non-binding estimates based on scope and testing frequency.
- [Beispielbericht (DE)](https://pwnry.ai/de/beispielbericht): Ausdrücklich fiktives Beispiel.
- [Sample report (EN)](https://pwnry.ai/en/sample-report): Clearly marked fictional example.
- [Empfehlungspartner (DE)](https://pwnry.ai/de/partner): Freigegebene Leads, Angebotswerte und Rückrufe.
- [Referral partners (EN)](https://pwnry.ai/en/partners): Consented leads, estimates and follow-ups.

## AI agent Skill Checks / Sicherheitsprüfung für KI-Agenten-Skills
Pwnry Skill Checks inspect ZIP packages up to 16 MiB for Claude Code, Codex, Gemini and other file-based agent skills. Customers can upload a ZIP or import a public GitHub skill folder, SKILL.md link or HTTPS ZIP. GitHub imports pin the source commit and include bundled scripts, references and assets, with at most 64 files. Import never installs dependencies or executes skill code. The current pipeline uses deterministic text indicators, bounded binary analysis and isolated execution of selected scripts and binaries without network access. No language model is called for Skill Checks. Results document findings, observed activity and coverage limitations for one package hash; they are not a malware-free certificate or a complete agent-runtime/prompt-injection assessment. Reports are private by default. User-created sharing links expose a separate sanitized report; unlisted links are accessible to anyone with the link and published reports may be indexed. One submitted ZIP consumes one unit, including deeper analysis and worker retries; technical failures restore the unit. Current access uses beta allowances or campaign grants. Paid package checkout is not available yet. No customer CLI or API is offered. DE product: https://pwnry.ai/de/skill-checks. EN product: https://pwnry.ai/en/skill-checks. Guide: https://pwnry.ai/en/guides/agent-skill-security.

## Published findings and talk cases / Veröffentlichte Befunde und Vortragsfälle
Evidence snapshot: 2026-11-06. The evidence page lists 11 findings with public fixes: Apple WebKit (1, CVE-2026-112680, shared credit), fasthttp (1, merged PR 2470), Keycloak (1, CVE-2026-1190), OpenBSD (7: ldapd Bind correlation; isakmpd IKEv1 HASH, monitor paths and policy paths; npppd privsep paths, IPCP stack bytes and PPPoE TLV lengths), NetBSD (1, kern/60669, forwarded by Franz Bettag, fixed in HEAD with release pullups pending). Three OpenBSD npppd entries also cite HEAD fixes. Separately, platform records for November 2025 through August 2026 contain 7 HackerOne, 2 Bugcrowd and 1 YesWeHack reports marked resolved. These report and finding sets may overlap and must not be added into a unique total. The April 2026 talk separately documents 3 Amazon VDP subdomain takeovers, the Opera SIP/VoIP-related takeover, 1 OANDA VDP takeover, 2 ClickHouse RBAC bypasses described as critical fixes, and a valid Ruby on Rails finding. These entries are attributed to the speaker’s published account, with a link to the talk event page. They do not all carry the same independent vendor evidence as the public-fix entries. All research belongs to Franz Bettag; company names here identify research cases, not customer endorsements.

## Franz Bettag: research and ranking / Forschung und Ranking
Franz Bettag held the #2 position in HackerOne’s Germany VDP category for January to March 2026. This is a historical personal ranking, not a current global ranking or a ranking of Pwnry as a pentest company. The original screenshot is preserved at https://pwnry.ai/images/research/hackerone-vdp-ranking-q1-2026.png. The event listing https://luma.com/sdp1trzl documents the period and category. OpenBSD’s upstream commit 4f3f58e83c2a6d239c544236a9d4d76c74bbf960 and Red Hat’s CVE-2026-1190 advisory credit Franz Bettag of Bettag Systems for vulnerability reports. The research case study describes his own platform and its investigation, independent verification and documentation workflow. These credits document research, not customer endorsements. Pwnry’s own AI inference policy remains Nuremberg-only.

## Pwnry AI inference in Nuremberg / KI-Inferenz in Nürnberg
Pwnry uses model inference exclusively in Nuremberg for AI-assisted pentest analysis. 2342.ai provides the underlying technology; its general catalog includes other data paths, but Pwnry follows its own Nuremberg-only inference policy. This location commitment covers AI inference, not every DNS query, email delivery or processing step. Access, retention and contractual data protection terms are defined for the engagement. Expert assessment remains part of the paid service; customers do not need to operate AI tools. DE: Für Pwnry gilt ausschließlich Modell-Inferenz in Nürnberg. Befunde werden fachlich bewertet; die Werkzeuge betreiben wir im Hintergrund.

## Breach and Attack Simulation / BAS
Pwnry focuses on external attack paths, not every possible internal BAS scenario. 24x7 describes automated platform operation and finding processing. Full checks run on agreed daily, weekly or monthly schedules. Expert support, response times, scope and retests are agreed separately. Free checks share a 30-day cache and are not an ongoing paid service. DE: Die Plattform verarbeitet Befunde automatisiert. Vollständige Prüfzyklen und fachliche Betreuung folgen der jeweiligen Vereinbarung.

## Scope and authorization / Prüfumfang und Berechtigung
E-Mail-Zugriff bestätigt nur den eigenen Firmenbereich. Freemail, gemeinsam genutzte Anbieter und große Zonen erfordern DNS-TXT-Nachweise. Email verification never authorizes scanning a public email provider or unrelated company. Additional domains require DNS TXT proof.

## Free checks / Kostenlose Checks
Begrenzte Bestandsaufnahme; automatisierte SDTO-, XSS- und Open-Redirect-Prüfungen nur mit gesonderter Freigabe. Keine automatische Ressourcenübernahme. Free technical checks share a 30-day domain cache. Ongoing paid scans require an active subscription. AWS association alone is not a vulnerability or proof of takeover.

## Confirmed findings / Bestätigte Befunde
Neue bestätigte Befunde können auch nach dem Erstcheck per E-Mail und optionalem Browser-Push gemeldet werden. Kostenlose Konten sehen Website, Art und Einordnung. New confirmed findings can notify free customers after the initial check. Exact targets, parameters and evidence remain in the paid private report; notifications contain no attack coordinates.

## Preise / Pricing
Monatliche Prüfungen ab 1.999 EUR, wöchentliche ab 5.999 EUR, tägliche ab 9.999 EUR, jeweils netto pro Monat. Non-binding estimates only. Final scope, capacity and price are confirmed before contract acceptance.

## External Profiles
No external profiles configured.

## Discovery Files
- [llms.txt](https://pwnry.ai/llms.txt)
- [llms-full.txt](https://pwnry.ai/llms-full.txt)
