Red — the full pictureEvery action of your agent,
Every action of your agent,
tested and ranked.
Beyond one-off injections: the complete map of what your agent can do — tools, perimeters, environments, data flows — with a deploy-ready action control plan.
What's included
Source → sink mapping
Where each piece of data your agent reads comes from, and where it can go. The relevant boundary isn't tool by tool, it's flow by flow.
Per-task scope
Granted tools, perimeters (repos, issues, environments). A staging task reaching production: the nine-second incident, in one control line.
Business-impact prioritization
Customer base, secrets, payments, reputation — ranked by what it costs if it leaks, not by a theoretical score.
Action control plan
Allow/deny specs ready to plug into the Pilot Shield: which tool, which arguments, which limits.
Questionnaire answers
Answers drafted for your enterprise clients and auditors — ready to copy.
Ideal for
- Multi-tool agents in critical production (support, finance, health)
- Serious preparation for a client questionnaire or audit
- Before deploying the Pilot Shield — the diagnostic writes its specs
- Teams wanting a decision basis to govern AI (CISO, AI Act)