Browser-local self-assessment / No login
Know the release boundary before the agent crosses it.
Identify control gaps before giving an AI agent production tools, credentials, data, or action authority. In 15 focused questions, map the controls that support a scoped release decision and generate a sample Permission Envelope.
Your profile, answers, score, and Permission Envelope stay in this browser. Nothing is sent unless you separately submit the optional contact form after your result.
release-control questions
readiness dimensions
priority gaps
result generation
Step 1 / Define the subject
Describe the release boundary—without sensitive data.
This profile shapes your sample Permission Envelope. Leave optional fields blank if the boundary is not yet defined.
Step 2 / Control assessment
Release-control maturity
“Not applicable” is not automatically safe. It receives limited credit because the assessment cannot confirm a control boundary that has not been demonstrated.
Step 3 / Self-assessment result
Dimension results
Scores are deterministic weighted completion values, not probabilities or safety guarantees.
Three highest-priority gaps
Priority combines the control’s maturity, its dimension weights, and the selected risk-bearing capabilities.
Suggested release restrictions
Use these as starting conditions. A responsible release owner should approve the final boundary.
Changes to the model, system instructions, tools, permissions, agent code, orchestration, retrieval sources, memory behavior, policy baseline, or deployment configuration can invalidate this result and require reassessment.
Sample Permission Envelope
This editable draft translates the assessment boundary into structured control fields. Review it before use.
Optional / Human scope review
Get help qualifying this release.
Send only contact details, your readiness category, high-level risk dimensions, and optional non-sensitive context. Your 15 answers and Permission Envelope are never included.
Do not include secrets, credentials, prompts, customer data, personal data, or confidential instructions.