Configuration and source analysis
Read-only review within the supplied scope. Record source provenance and avoid copying unnecessary sensitive values.
Defensive planning for authorized work
Plan the stop before the run starts.
Define permission, checkpoints, evidence, budgets, and recovery before an audit agent spends its first token.
Authorized run plan
Every required boundary stays visible before local simulation begins.
Scope and authority
State the permission boundary in plain language.
Run budgets
Stop before uncertainty consumes the remainder.
Modeled use0%
The run will pause before the review threshold.
Checkpoints
Save enough state to resume safely.
Evidence capture
Attach provenance without leaking secrets.
Evidence policy
Preview redaction locally before export.
Redacted output appears here.
Review sensitive steps before they block the run.
These categories plan review and evidence only. They never execute security actions.
Human review required
Any active validation must be separately authorized, manually reviewed, and performed outside this planner using approved organizational procedures.
Minimize and redact
Collect only what is necessary, use test data where possible, redact evidence, and follow the selected retention policy.
Never part of this run
Exploit execution, bypass attempts, destructive testing, persistence, credential attacks, evasion, and out-of-scope scanning are excluded.
Recovery should be a saved state, not a second investigation.
A useful checkpoint explains what was authorized, what evidence was captured, what consumed the budget, why the next step stopped, and exactly what a reviewed resume may do next.
Draft
Complete authorization to make the plan ready.
None recorded
Record one before sensitive or expensive work so a reviewed resume starts from known state.
Every decision leaves a receipt.
The export contains the plan, redacted evidence, checkpoints, budget snapshots, policy decisions, and state transitions.