How to turn an AI skill into a shared internal tool
Turn a repeated support skill into a tool your team can use: define the inputs, put fixed steps in code, bound AI decisions, and review sharing and access.
AI workflow guides
Practical guides to turn AI skills into shared internal tools, reduce repeated model work, and keep control of actions and access.
A support-ticket triage workflow your team can review
A worked Zendesk ticket example with a category table, queue suggestion, reply template and human-review rules. Start with pasted text and a copyable result.
Turn Jira notes into a weekly Slack update
A Jira-to-Slack workflow blueprint: three example tickets, status counts, visible blockers and a copyable update a person reviews before posting.
How to reduce token usage in repeatable AI workflows
Put repeated steps in code, keep AI for judgment, and compare equivalent runs. A worked token ledger includes setup, retries, quality checks and break-even.
The lethal trifecta is in your support inbox
Support, sales and recruiting routines read text strangers write, see private data and send messages out. That is everything prompt injection needs to leak data. One attack, end to end, and the fix.
Why detection can’t stop prompt injection
Heuristic defences give no formal guarantee, and adaptive attacks get through them. Fixed plans change that, at a cost worth knowing.
When you don’t need Cresec
Cresec is for routines with all three legs of the lethal trifecta. If yours has fewer, or every step is already fixed by hand, you probably don’t need us. A candid guide.
How to share an AI workflow with your teammates (Claude, ChatGPT, Lovable, Base44, v0)
You built a routine, a skill, a GPT or a small app and your teammates want it. Here is how to share each kind — and what to do when it needs their own email, calendar or CRM.
Why IT blocks the AI tool you built — and what they are actually worried about
Your homemade AI tool works. So why does IT keep saying no? The lethal trifecta behind the “no”, the four questions that follow it, and how to answer each one.
Shared logins and AI tools: the hidden cost of one account for the whole team
Sharing one account is the fastest way to give your team an AI tool. Here is why it causes trouble later, and what to do instead.
Draft first: a safer way to let AI act on your behalf
Why the safest AI tools start by preparing drafts instead of sending, and how to move from drafts to sending one person at a time.
How to get IT to approve your AI tool: a one-page checklist
A short checklist to fill in before you ask IT or security to approve an AI tool you built for your team, starting with the lethal trifecta and the complete list of what the tool can do.
What is a governed AI tool?
A plain-language definition of a governed AI tool: code decides the action, with clear identities, bounded permissions, attributable records and explicit failure behavior.