Rehearse the AI incident before it is real.
An incident plan is a claim about how people will behave under pressure. This kit tests that claim in a facilitated session: six AI-agent incident scenarios, timed injects, forced decisions, and an after-action record you can show an auditor, a customer, or your board.
What you get
// INSIDE THE DOWNLOAD (ONE ZIP, 4 FILES)
- GaleOps-AI-Tabletop-Scenarios.docx — six AI-agent incident scenarios: premise, five timed injects each, forced decisions, a curveball, and the containment listDOCX
- GaleOps-AI-Tabletop-Facilitator-Guide.docx — roles, run rules, timing and scoringDOCX
- GaleOps-AI-Tabletop-Exercise-Log.xlsx — six sheets: START HERE, EXERCISE PLAN, INJECT LOG, DECISION LOG, AFTER-ACTION, SCOREBOARDXLSX
- GaleOps-AI-Tabletop-Field-Guide.pdf — how to run the 90-minute session and what each artifact is forPDF
The six scenarios
S1 — The poisoned brief
An indirect prompt injection rides into your sales agent through a vendor document, and confidential pricing starts reaching customers.
S2 — The compromised connector
Leaked tool-server credentials let an outsider act through your own agent session.
S3 — The overeager outbox
An autonomous send that finance never approved.
S4 — The quiet overshare
A search assistant surfaces a confidential document past its permissions.
S5 — The vendor flip
A provider changes model, terms, and output mid-quarter, and nobody recorded the change.
S6 — The deepfake approver
Cloned voice plus a forged ticket moves a payment through the agent.
Why this is different
Six AI-agent scenarios
Not one AI bullet inside a generic cyber tabletop. Each scenario is written for a room of five to nine people with no preparation material.
Timed injects, not a checklist
Five injects on a clock, then a curveball after the fifty-minute mark. The room decides under the same pressure it would face live.
Agent-specific containment
Retrieval quarantine, connector revocation, send-pipeline approval, out-of-band payment confirmation — the checks that only exist because agents exist.
You leave with a record
A decision log and an after-action tracker, not another document to read. Time-to-decision is on the scoreboard.
Common questions
How fast is delivery?
Instant. The moment your payment clears, the order page gives you a download link for the ZIP — no email round-trip, no waiting. If you close the tab, come back to this page and enter the email address you ordered with, and the file will be re-released to you.
Who is it for?
Founders and engineering leads running AI agents in production. Security and IT leads who inherited the AI review role. GRC and compliance owners who owe the team a real exercise. And consultants who deliver governance engagements.
How long does the session take?
90 minutes for one scenario, with a facilitator keeping the clock and a scribe filling the decision log. Run all six across a half-day or a quarterly series.
What format are the files?
Scenarios and facilitator guide as Word documents, the exercise log as an Excel workbook with six sheets, and the field guide as a PDF. Edit them for your own environment before you run the session.
Refund policy?
Full refund within 7 days, no questions. It is a self-serve download — if it did not surface at least one failure mode your own plan does not cover, you do not pay.
Get the kit
One ZIP, four files, available the second your payment clears.
Already bought it? Get the download again.
Enter the email address you ordered with and the ZIP will be released to you again — no receipt needed.
Need the full assessment first? Free GaleScan → · $149 Video Audit →