InsideAI Agent Attack Surface: Every Pass Through the Loop Is a Way In
How an AI agent differs from a chatbot, its 5 parts, the ReAct loop and its input and output channels, plus a lab to fingerprint an agent over its API.
[01] Recon — who you have found
The offensive security notes
I wish someone had left me.
I'm Bikramjeet Singh, a Lead Penetration Tester. Every engagement teaches something the courses skip over, so I write those bits down in full, web, API, Active Directory, cloud, Kubernetes and AI, including the attempts that went nowhere. No theory, no vendor filler, no "it depends".
Latest — the newest write-ups
What I learned recently
Newest first. Each one starts with something I did not understand, works through what actually happens under the hood, and ends with what it means for security. The wrong turns stay in.
AKIA or ASIA? How Entra ID, AWS and Google Cloud Log You In, and Where the Keys Hide
How Entra ID, AWS and Google Cloud log people and apps in, where keys and tokens hide on a test, how to reuse them in a lab, and what to check on each.
Client Sign-Off Isn't Enough: Cloud Pentest Rules for AWS, Azure and GCP
What AWS, Microsoft and Google let you test in a cloud pentest: permitted services, banned attacks, approval forms, reporting windows and a scoping checklist.
A Compression Library Mistook Bad Data for Send More, and Node.js Got an RCE
CVE-2026-111497 (Node2Shell) is a CVSS 9.2 Brotli decoder overflow in Node.js, now in CISA KEV. Learn the root cause, who is exposed and how to check and patch.
[02] Enumerate — map the attack surface
Eight tracks. Five are live, three are next.
Everything I publish lands on one of these. The count is the honest depth, not a target.
S0142
S0230
S0309
S0403
S0501
S06
S07
S08
Kubernetes Journey
Cluster security learned in public, start to finish. The series is complete.
AI / LLMs
Attacking and securing AI systems and the plumbing around them.
Threats & Exploits
Fresh CVEs pulled apart to root cause, not "patch now".
Cloud
IAM, storage exposure and config reviews that matter.
Web & API
App and API testing, auth flaws and business logic abuse.
AD & Infrastructure
Active Directory, lateral movement, privilege paths.
Coming soonSystem Breakouts
Kiosk escapes, restricted shells, thick client, mobile.
Coming soonWireless
WPA handshakes, evil twin, deauth, Bluetooth and RFID.
Coming soon Playbooks — the notes you run on each surface
The operator notes, by track.
Every live track has a command-level playbook. The real checks, the exact commands, what a hit looks like and the honest gotcha, ordered the way a test runs from recon to cleanup. Each note stands on its own, there's a tick-box checklist you can run mid-engagement, and live search across the lot. Cloud sorts by AWS, Azure and Google Cloud.
N01254
N02140
N0338
N0426
N05
N06
N07
Open the Pentest Notes command-level operator playbooks ▸
AI / LLMs
The OWASP LLM path: prompt injection, jailbreaks, data and output handling, agents and tools.
Kubernetes
The full attack path: recon, in-pod enumeration, escalation, escape, cleanup, plus the scanner and CIS tooling.
Cloud
Sortable by AWS, Azure or Google Cloud, scoping and recon all the way down to cleanup.
AWS · Azure · GCPWeb & API
Recon to reporting: auth and sessions, access control, injection, SSRF, client-side, API-specific.
AD & Infrastructure
Active Directory, lateral movement and privilege paths.
Coming soonWireless
WPA handshakes, evil twin, deauth, Bluetooth and RFID.
Coming soonSystem Breakouts
Kiosk escapes, restricted shells, thick client and mobile.
Coming soon [03] Escalate — the receipts
None of this is theoretical
Every line below is a link to something already on this site. No roadmap, no coming soon, just what is published and how much of it there is.
On the shelf right now
E01Kubernetes Journey 42 topics
E02A complete series: a cluster built, broken and rebuilt across 42 topics, from RBAC and admission control through to runtime.
AI / LLMs Journey 30 topics
E03How the models actually work, then how they fail, from neural network internals to attacking LLM apps.
Cloud Journey 3 topics
E04Testing AWS, Azure and Google Cloud from the ground up, starting with what each service is really called.
Pentest Notes 458 notes
E05Command-level operator playbooks: 254 AI / LLMs, 140 Kubernetes, 38 Cloud, 26 Web & API. The bits you reach for mid-engagement.
Threat briefs 9 briefs
CVE teardowns taken to root cause rather than "patch now and hope".
Verified
OSCPOffSec
PNPTTCM Security
CASAApp Security
CKSKodeKloud course
Findings mapped to
OWASP OTGMITRE ATT&CKCWESANS Top 25CIS Benchmarks
Counted honestly
85write-ups
~18hreading
458notes
5live tracks
Started the Kubernetes track in the open and kept going for 42 topics. The counts move because I publish, not because I planned a number.
[04] Report — hand it over
If it was useful, tell me. If I got it wrong, tell me faster.
New write-up most days. Labs, dead ends and the odd working exploit. Compare notes, argue with me, or just come along for the ride.