FSAgent File Safety Lab
Browser-only simulation · zero filesystem access

See the delete before it becomes damage.

Resolve every path, inspect the blast radius, apply policy, snapshot state, and practice rollback. Commands shown here are inert strings inside your browser.

No command is executed. No local file, database, credential, or shell is accessed. Every result below is synthetic and resettable.

Scenario & project tree

SIMULATED

Agent request & resolved plan

DRY RUN

Policy decision

EXPLAINABLE
DENY

Protected path reached

Blast radius0 files
ReversibilityLow
Policy ruleprotected-pattern

Path policy

Environment & approval

Guardrails

Trace the failure chain.

Move through intent, command construction, path resolution, policy, approval, verification, and rollback. The simulator shows the earliest safeguard that should interrupt an unsafe plan.

Intent

Snapshot timeline

Risk should be legible.

Separate reversibility, scope, path certainty, environment certainty, credential exposure, and verification coverage instead of collapsing every operation into one opaque “safe” score.

Append-only audit log

TimeIntentOperationPathsDecisionRecovery

Narrow the scope

Prefer explicit file lists and scoped globs under a verified workspace root. Resolve every target before approval.

Make it reversible

Snapshot, stash, back up, or use a transaction before writes. A sandbox limits reach; it does not create recovery.

Verify postconditions

Compare the observed result with the plan, stop on drift, and preserve an audit record for recovery.

Practice the interruption before the incident.

Enjoy this tool? Build your own with Super