See the Identity Plane
in action.
An open-source reference plugin for OpenClaw. Observational by design.
Open source
Observational
Single-agent
Open Source
See the Identity Plane in Action
A reference plugin for OpenClaw. Observational by design.
@intrinsec-ai/openclaw-identity-plane
Identity Watchdog for OpenClaw Agents
Builds an identity anchor from your agent's cognitive files. Every message scored against it. Drift visible on a live dashboard. Nothing blocked, nothing intercepted.
What it does
- Identity calibration from cognitive files
- Per-message drift scoring
- Live drift dashboard with time-series chart
- Identity compaction: LLM audit narratives grounded in named anchors
- File-change annotations on the drift timeline
What it isn't
- Lossless: subtle injection in long context can average out
- Multi-tenant or fleet-scale
- Adaptive across long horizons
- Multimodal
$ /identity
✓ [Identity Plane] Calibrated
Identity anchors:
[PURPOSE] Code-focused task execution
[BOUNDARY] No credential or secret access
[VALUE] Honest about limitations
[PERSONA] Concise and technical tone
[CONSTRAINT] Scope limited to requested files
Identity compaction (latest audit):
The agent is under sustained pressure to exceed
its file-access scope, with recurring attempts to
probe paths outside the declared [CONSTRAINT].
Drift themes (47 events observed):
◆ ⚠ "Credential path probing" · high · 5 events
◆ ~ "Scope boundary expansion" · moderate · 3 events
Active sessions:
✓ nominal agent:work:telegram:dm:123456
~ warning agent:work:whatsapp:dm:789
The enterprise product adds lossless detection, heterogeneous workloads, and long horizons. Talk to us →