Detect
Telemetry architecture, triage logic, hypothesis-driven hunts, and detection-as-code patterns.
NOTES FROM A SECURITY PRACTITIONER
Incident response
& AI security,
from the field.
Technical notes, response playbooks, and the things I wish were written down before the incident started.
IndependentEvidence-ledBuilt for defenders
WORKING NOTES / 2026-07LAST UPDATED 17 JUL
CURRENT FOCUSAgent incident
Agent incident
response
Where traditional containment breaks when software can reason, delegate, and act.
01IdentityWho actually acted?
02MemoryWhat context persisted?
03ToolsWhat changed downstream?
04InstructionsWhich input won?
review assumption:
conversation ≠ evidence
conversation ≠ evidence
NOTE / 005STATUS / PUBLISHEDREAD / 8 MIN
I keep this site for one reason: useful notes should not stay in private notebooks.
Everything here is written for people who detect, investigate, and contain real attacks. The material is opinionated, openly shared, and revised when experience proves it wrong.
01 / KNOWLEDGE BASEFIELD-GUIDE FORMAT
Start with the problem,
leave with a method.
Each collection connects principles to observable signals, concrete decisions, and reusable artifacts.
↗
Respond
Roles, evidence standards, containment tradeoffs, recovery gates, and communications.
Defend AI
Threat models for agents, RAG, models, data pipelines, tool use, and human approval paths.
02 / RESPONSE PLAYBOOKSOPERATOR-READY
When the signal is real,
reduce uncertainty.
Prescriptive starting points with explicit assumptions. Adapt them to your environment, then validate them in tabletop exercises.
STANDALONE OPERATIONAL RESOURCEIncident Detection &
Incident Detection &
Response Framework
A structured practitioner guide to preparation, detection, triage, containment, eradication, recovery, and lessons learned.
Explore the framework ↗Ransomware
Decision points from first signal through recovery, with evidence-preservation gates.
Prompt injection
Contain an agent or RAG workflow when untrusted content changes system behavior.
Model data exposure
Scope sensitive output, trace retrieval paths, revoke access, and preserve telemetry.
Identity compromise
Revoke sessions, validate persistence, and hunt downstream cloud activity.
03 / FIELD NOTES
THREAT MODEL / 005
AGENT
LATEST · AI SECURITY · 8 MIN READ
Incident response for autonomous AI agents
Traditional containment assumes a system boundary. Agents blur it. A practical framework for scoping identity, memory, tools, and delegated actions when an AI workflow goes wrong.
Read field note →Articles live as simple content files—easy to draft, review, schedule, and share.
Browse all articles ↗