Status: reviewer-facing ecosystem note.
This document explains how CML fits into the broader trustworthy-agent evidence architecture.
Central map:
https://github.com/safal207/L-THREAD-Liminal-Thread-Secure-Protocol-LTP-/blob/main/docs/ECOSYSTEM_SPIDER_MAP.md
CML is the causal permission and responsibility lineage layer.
It asks:
Why was this action allowed?
and checks whether the causal parent, approval, intent, or responsibility chain is intact.
PythiaLabs — pre-execution evidence gates for high-risk agentic actions
LTP — path-level trace/replay/admissibility protocol
CML — causal permission and responsibility lineage
DMP — decision memory and irreversibility governance
LRI — living identity and relational invariants
| Layer | Relationship |
|---|---|
| PythiaLabs | CML provides causal permission signals that can help an action gate allow, block, or escalate a proposed action. |
| LTP | LTP preserves the path-level trace where CML causal validity can be inspected over time. |
| DMP | DMP preserves decision memory when a causally permitted action later creates durable or irreversible consequences. |
| LRI | LRI adds human identity and revisability boundaries when causal permission is not enough to protect the person. |
CML should not be read as an ordinary logging tool.
Logs show what happened. CML checks why it was allowed.
CML does not claim:
- full AI alignment;
- production security certification;
- regulatory compliance certification;
- replacement of logs, tracing, policy engines, or human review;
- universal prevention of unsafe actions.
The narrower claim is:
CML is an open-source causal audit layer for checking permission and responsibility lineage in structured action traces.