Hey there. This past week was a whirlwind. Last week, I got the itch to start hacking and went on a pretty intense spree all weekend. I’ve been dialing in my automation and started finding some great issues using my autonomous agent system. I didn’t want to lose momentum, so I prioritized that over getting the newsletter out last week. Sorry about that.
I may have overdone it though, because I ended up catching a pretty bad cold and it knocked me out for most of this week. Anyway, I’m back now and almost fully recovered, so I’m hoping for a better week ahead.
Looking at the news, there isn’t a ton to report, but we do have a handful of cool write-ups to highlight. Let’s dive in.
I’m available for 1:1 calls if you want to chat about bug bounty, career growth, community building, or anything else you think I can help with. You can book time with me here.
Inside a Large Self-Hosted VRP (Sam Erb) [🎥 Video]
by Bug Bounty Village
Sam Erb breaks down how Google/Alphabet runs a large self-hosted VRP, covering triage at scale, escalation paths, payout dynamics across overlapping programs, and why CVSS isn’t central to their internal process. The technical highlight is a deep dive into GZIP/Deflate manipulation (LZ77 + dynamic Huffman quirks) used to reach script injection on a google.com subdomain, including language-level differences between Go and Python gzip implementations.
Have something you want to Spotlight? Tell me.
Burp Extension Detects Maps/Gemini API Key Leaks [𝕏 Tweet]
by Nathan Jones (@njcve_)
A tweet announcing a Burp extension that scans traffic for exposed Google API keys and validates whether they provide access to Gemini/related APIs. The post links to the repo and frames verification as the key step for reducing false positives during key-leak hunting.
CSP Analyzer Flags Risky Sources and Common Bypass Candidates [𝕏 Tweet]
by payloadartist
A tweet pointing to a CSP analysis site that parses a policy and highlights potentially abusable allowances, including risky third-party script origins. It’s a quick way to sanity-check CSP posture and identify where a bypass might be feasible in real-world apps.
VibeSec-Skill: Security-Focused “Skill” for AI Coding Assistants [🛠️ Tool]
by Behi
An open-source “skill” pack aimed at steering AI coding assistants toward secure patterns across common bug bounty classes (IDOR, SSRF, XSS, CSRF, auth/JWT issues, unsafe file upload, and more). The repo is structured as reusable checklists and guardrails intended to reduce vulnerable output and push for explicit validation and edge-case handling.
Trafexia: Electron-Based Mobile Traffic Interceptor [🛠️ Tool]
by danieldev23
An Electron desktop proxy tool for capturing and replaying mobile app HTTP(S) traffic, with a request composer, replay support, and timeline-based browsing. It’s built for practical mobile debugging and vulnerability reproduction, with filters for quickly isolating interesting requests and responses.
S3-Checker: Concurrent S3 Bucket Discovery and Exposure Classification [🛠️ Tool]
by zapstiko
A Go-based CLI that generates likely bucket names, checks existence, and classifies exposure signals (public vs. private) with parseable output for pipelines. It also supports optional OSINT enrichment (e.g., GrayHatWarfare) to speed up triage when buckets appear in external indexes.
Have a favorite tool? Tell me.
Vibe-Hacking Cloudflare's vinext (LLM-Generated Next.js Replacement) [📓 Blog]
by Hacktron AI (@HacktronAI)
This audit of Cloudflare’s vinext reports dozens of findings, including session mix-ups from AsyncLocalStorage fallback behavior, cache poisoning caused by incomplete cache-key construction, and routing/middleware normalization differentials that lead to auth bypasses. The write-up includes concrete PoCs and root-cause analysis, underscoring how “works in tests” LLM output can still collapse under adversarial edge cases.
Google API Keys Weren't Secrets. But then Gemini Changed the Rules. [📓 Blog]
by Truffle Security
Truffle Security argues that Google API keys are effectively becoming higher-value secrets as Gemini-enabled capabilities increase the impact of key exposure (abuse, unexpected billing, and broader access). The post focuses on practical mitigations—tight key restrictions, rotation, and CI/CD secrets scanning—and what security teams and bug bounty triage should change as harvesting becomes easier to automate.
Python Quirks That Turn Into Path Traversal and RCE Primitives [𝕏 Tweet]
by YesWeHack
YesWeHack highlights research showing how surprising behavior in common Python utilities (e.g., path and URL joining, unsafe deserialization) can be turned into traversal and, in some cases, code execution. It’s a reminder that “safe” helper functions can become exploit primitives when assumptions leak into security boundaries.
OAuth Popup Iframe Hijack via Predictable window.open Name [📓 Blog]
by Critical Thinking Podcast
This research shows an OAuth flow abuse where a hardcoded window.open target name enables iframe/popup hijacking. By pre-positioning a same-origin frame and steering it into the OAuth callback context, the attacker can force unwanted account/app linkage and leak workspace metadata in downstream flows.
Total.js RCE Gadgets Across TextDB, FlowStream, and Component Installs [📓 Blog]
by Critical Thinking Podcast
A deep technical breakdown of multiple RCE paths in Total.js v4/v5, centered on new Function()-backed compilation in query logic and FlowStream components, plus command injection during component npm installs. The post includes line-level references, bypasses for naïve blacklists, and working PoCs for both blind and non-blind execution.
This write-up details a chained attack against OpenClaw (formerly Moltbot/ClawdBot) culminating in a single-click RCE (CVE-2026-25253). It walks through the kill chain from a user visiting a malicious page to code execution in the agent context, then shows how over-privileged “god mode” behavior enables theft of tokens and local secrets.
Hackian and the Ghosts in Your Business Logic [📓 Blog]
by Ethiack (@Ethiack)
Ethiack introduces “Hackian,” an agent positioned to find business-logic failures by simulating workflows across multiple identities rather than scanning for single-endpoint bugs. The post includes two example disclosures: SonarQube branch metadata exposure due to defaults, and an unauthenticated n8n settings endpoint leaking configuration/infrastructure metadata (fixed in v1.118.0).
Did I miss something? Tell me.
Hacking OAuth: How Token Theft Enables Account Takeover [𝕏 Tweet]
by Bugcrowd
Bugcrowd shares a short primer on OAuth abuse patterns where token theft can translate into account takeover without passwords. The thread frames common failure modes in “Login with X” implementations and points to a deeper breakdown of the flow and attack surface.
What I Learned Building AI Agents for Bug Bounty Hunting [📓 Blog]
by Bugcrowd
A guest post on building a multi-agent pipeline to discover and triage sensitive documents indexed by search engines, using separate agents for discovery, scraping, classification, and report drafting. It emphasizes operational lessons—false-positive control, keeping critical logic in deterministic automation, and using LLMs where they add leverage (summarization and reporting).
How I Cured My Bug Hunting AI's Goldfish Memory [📓 Blog]
by labs.trace37
This post proposes a hook-based architecture for stateful, auditable AI-assisted bug hunting, with lifecycle scripts for session start, tool gating, post-tool learning, retry detection, and handoff. The focus is on controlling agent behavior (confidence gates, structured logs, and clean handovers) rather than model prompts or payloads.
A research repo exploring how well-known technique names and short payload lists can create blind spots for code injection and SSTI testing. It collects variations and methodology meant to push beyond canonical examples and surface edge-case behaviors that basic cheat sheets miss.
Did I miss something? Tell me.
Inside H1-65: OKX’s Live Hacking Event in Singapore [🎥 Video]
by HackerOne
A recap of HackerOne’s H1-65 live hacking event with OKX, focused on the format and collaboration dynamics rather than exploit details. It highlights coordinated testing and vendor–researcher feedback loops, but stays largely non-technical.
CTBB Podcast Q&A: HackerOne Answers Community Questions [𝕏 Tweet]
by Critical Thinking Podcast
A tweet linking to a CTBB Q&A episode featuring HackerOne, framed around responding to community questions and collecting more. The content is positioned as platform/program discussion rather than a technical exploit walkthrough.
I Hacked My First AI Chatbot [🎥 Video]
by Ben Sadeghipour (@NahamSec)
NahamSec walks through two CTF-style chatbot challenges, focusing on prompt injection and context manipulation to extract protected data and flags. The video illustrates practical testing loops for distinguishing trusted vs. untrusted model context and how reflected input can override system intent.
This Is What Real Bug Bounty Recon Looks Like [🎥 Video]
by Logan-sec (Logan-sec)
A long-form recon session showing manual mapping of a real target’s workflows, focusing on how object identifiers and operational flows can translate into IDOR-style opportunities. The emphasis is methodology and note-taking over tool-driven scanning.
How Hackers Technically Unlock Paid Apps [🎥 Video]
by DeadOverflow (deadoverflow)
DeadOverflow covers common reverse-engineering approaches used to bypass paid-app restrictions, including patching client-side checks and tampering with app logic. The discussion is framed around defensive implications, emphasizing server-side validation and tamper resistance over relying on local enforcement.
Will AI Replace Bug Bounty Hunters??? (WYBUT #3) [🎥 Video]
by Harrison Richardson (@rs0n)
A discussion episode on where AI currently helps (recon acceleration, triage support, drafting) and where it still fails (novel exploitation and adversarial reasoning). The conversation focuses on how hunters can adapt by treating AI as throughput leverage while keeping manual validation and creative research central.
Open Redirect Bypass via Hostname Prefix Confusion [🎥 Video]
by Gospel
A PoC showing how naïve hostname validation (“starts with domain.app”) can be bypassed with a lookalike attacker-controlled domain (e.g., domain.appattacker.com). The video reinforces why redirect allowlists must use proper URL parsing and exact host matching, not substring or prefix checks.
Did I miss something? Tell me.
Gemini on Android Reportedly Exposes a Tool Call That Reads Live Notifications [𝕏 Tweet]
by Critical Thinking Podcast
A report that Gemini on Android exposes tool calls not present on the web client, including one that can read live notifications (including SMS). The post notes a $1337 bonus from Google, highlighting the security implications of platform-specific capability gaps in agent/tooling surfaces.
IDOR Hunting Workflow: Map IDs, Mutate One Dimension, Test Non-GET Paths [𝕏 Tweet]
by André Baptista (@0xacb)
A compact IDOR workflow focused on systematically mapping object IDs across endpoints and changing one dimension at a time to spot inconsistent authorization. It also calls out “soft failures” (partial responses/empty arrays), non-GET methods (PATCH/DELETE), and cross-client inconsistencies between UI, API, and mobile.
XXE Filter Bypasses: External DTDs, Parameter Entities, Encodings, and Wrappers [𝕏 Tweet]
by Intigriti
A checklist of XXE evasion ideas, including external DTDs with parameter entities, alternate encodings to dodge keyword filters, and wrapper tricks in PHP ecosystems. It also flags second-order XXE as a recurring blind spot when XML is processed asynchronously.
Redirect Parameter Checklist: Scheme Smuggling, Host Confusion, and Chaining [𝕏 Tweet]
by André Baptista (@0xacb)
A set of redirect-parameter tests covering scheme smuggling, userinfo/host confusion, and encoding/double-decode behavior. It also ties open redirects to higher-impact chains like OAuth token theft and SSRF when redirect destinations feed backend fetchers.
Hidden Parameter Enumeration: Crawl, Mine JS, Bruteforce, and Dork [𝕏 Tweet]
by Intigriti
Intigriti summarizes common ways to surface undocumented parameters, including crawling, extracting client-side parameter usage, JavaScript analysis, bruteforcing with tailored wordlists, and search-based dorking. The emphasis is on finding injection surface that never appears in UI-driven testing.
Did I miss something? Tell me.
Because Disclosure Matters: This newsletter was produced with the assistance of AI. While I strive for accuracy and quality, not all content has been independently vetted or fact-checked. Please allow for a reasonable margin of error. The views expressed are my own and do not reflect those of my employer.