Agent design · Founder-curated
scope baseline
A four-question template that pins agent scope before any tool wiring.
informationalassisted workaction takingpurpose and scope
Last reviewed: 2026-05-10🔒 Full guidance in workspace
Reference library
The AgentProof Reference Library
A curated, source-backed library of agentic AI readiness guidance — indexed by topic, capability zone, control relevance, and source family. Public preview shows the index; full guidance is available in your workspace.
Source-backedLast-reviewed datesCapability-zone taggedPublic preview
Browse the library
Filter by topic, source family, and capability zone. Filters become interactive in the workspace.
Topic
Agent designHuman oversightData sensitivityAction authorityTesting and evidenceMonitoring and reassessmentMicrosoft-style environmentsAI governance signals
Source family
StandardsVendor docsResearchFounder-curated
Capability zone
informationalassisted workaction taking
Reviewed
Human oversight · Founder-curated
oversight patterns
Patterns for human-in-the-loop by capability zone. Public preview is a summary.
assisted workaction takinghuman oversight
Last reviewed: 2026-05-12🔒 Full guidance in workspace
Data sensitivity · Standards
data classification
How to classify what an agent reads — and when classification belongs at the data layer.
assisted workaction takingdata access and sensitivity
Last reviewed: 2026-04-22🔒 Full guidance in workspace
Action authority · Standards
action authority
Allow-list discipline + idempotency for action-taking agents.
action takingaction authority
Last reviewed: 2026-04-28🔒 Full guidance in workspace
Testing and evidence · Founder-curated
evidence minimums
What an audit-safe action trace must capture; the full checklist lives in the workspace.
informationalassisted workaction takingtesting and evidence
Last reviewed: 2026-05-03🔒 Full guidance in workspace
Monitoring and reassessment · Research
monitoring baseline
Defining a 'normal' baseline before go-live so drift is detectable.
assisted workaction takingmonitoring and reassessment
Last reviewed: 2026-04-15🔒 Full guidance in workspace
Microsoft-style environments · Vendor docs
microsoft style
Plain-English summary of public Microsoft agentic guidance. AgentProof does not speak on behalf of Microsoft.
assisted workaction takingdata access and sensitivityaction authority
Last reviewed: 2026-05-08🔒 Full guidance in workspace
AI governance signals · Research
governance signals
Tracking governance signals that change what 'ready' means for agentic AI workloads.
informationalassisted workaction takingpurpose and scopemonitoring and reassessment
Last reviewed: 2026-04-20🔒 Full guidance in workspace
Public preview shows
- Entry title, topic, and source family
- Public-safe excerpt
- Capability-zone + control relevance tags
- Last-reviewed date and review status
Workspace unlocks
- Full guidance text with source citations
- Stale-source warnings on affected reports
- Direct deep-links from report findings
- Improvement-cycle suggestions wired to each entry
- Reassessment trigger when an entry's source changes
How AgentProof keeps this current
Every entry is owned. Every change passes a human review gate. Every pack release notes which entries changed and which reports are affected.
Open the AI landscape radar →Ready to assess your own agent?
Start the free assessment to apply this guidance to a real agent. No payment. No public registration required.