Strategy. Design. Development.

Branding and websites for ambitious businesses.

From Nairobi to the world.

Digital marketing that connects to sales.

Your goals. Our craft. Real results.

Strategy. Design. Development.

Branding and websites for ambitious businesses.

From Nairobi to the world.

Digital marketing that connects to sales.

Your goals. Our craft. Real results.

Template

Agentic competitive research threat card

Every door a web-browsing research agent can be pushed through, the control that closes it, and the kill switches to test before the first run. No control, no launch.

View as Markdown

Fill one card per agent. If a row has no control, the agent does not run. Write NONE and stop. Do not write “the prompt tells it not to.”

Test every kill switch before launch. A cap that has never fired is a cap nobody tested.

Drawing diagram…

The agent reads public pages and writes notes to one store. Everything else is a person or a script.

Is an agent the right shape?

If you can name the pages, use a scheduled fetch, a diff, and one model step instead. An agent is for open questions where the path is not known.

Toolset (three tools, no more)

ToolAllowedLimit written here
Search public webYesQueries per run: ____
Fetch page (GET only)YesPages per run: ____ Bytes per page: ____
Write note to notes storeYesOne store, append only, schema-validated
Send email or post to chatNoDelivery is a separate approved step
Run code or shellNoNot needed
Access the customer relationship management (CRM) system, billing, or filesNoNo credentials in the agent environment

Budgets enforced by the runner

Doors and controls

DoorControl (write it)Tested on
Injected instructions in a pageDate
Scope creep (logins, forms, private areas)Date
Fabricated findingsDate
Runaway cost or loopDate
Leaky context in queries or URLsDate
Quiet drift after a prompt or model changeDate

Output gate

Kill switches (test each one)

Change control