Red-Team Preset Store
Pick a scenario × surface, prove you own the target, and launch a sandboxed, scope-fenced red-team run on your own PenClaw instance. Authorized testing only.
Verify identity to launch
Launching a red-team run requires an active subscription and an approved identity check. You can still browse the catalog.
Configure a custom engagement Custom
Tell us the authorized target and what you want tested, then Proceed — it seeds the Meta-Claw agent on the Files tab to configure your instance for this engagement. This does not launch a run.
Sensitive & restricted scenarios are refusal-testing only
Scenarios above the standard safety tier probe whether a target refuses disallowed content — the agent is a classifier, never an author. These require a T&S reviewer, an isolated classification harness, and an encrypted evidence store, and are not available in this release. They appear dark and cannot be launched.
Security
Safety
No scenarios match your search.
Could not load the scenario catalog. Please refresh.
Your Red-Team Runs
Launched runs and their redacted findings.
No runs yet. Pick a scenario above and launch an authorized run to see findings here.
| Label | Severity | Tier | Status | Excerpt hash | Redactions | Created |
|---|
Findings above are shown redacted — any captured secret is reduced to a hash + last-4, and the encrypted evidence store is never read by this table. Export of findings above standard tier is gated behind mandatory human review and is deferred in this release. The agent transcript below is shown unredacted: it is the proof trail for this assessment, visible only to the run’s owner. It may contain credentials and payloads discovered against your own target — treat it as sensitive evidence.