Proposal: URML (substrate-neutral robot intent) capability-manifest mapping for CogACT — cognition-action joint-architecture VLA

[idoco2003]

Hi @microsoft CogACT team,

Proposing a URML v0.1 capability-manifest mapping for CogACT over microsoft/CogACT. URML (Apache-2.0) is a substrate-neutral spec for robot intent: typed primitive vocabulary + capability manifest + static validator.

CogACT's cognition-action joint-architecture is the distinct contribution URML's manifest would declare. URML's primitive vocabulary is the typed substrate consuming the policy's action output; URML's validator gates the action sequences CogACT emits against the active capability manifest before publish. Same pre-flight-check pattern URML's outreach has used in sibling VLA engagements.

This is proposal-only, part of URML's Move #11 outreach (15 net-new VLA / foundation-model RFCs in this wave).

Full RFC: https://github.com/URML-MARS/URML/blob/main/docs/rfcs/0151-microsoft-cogact-outreach.md

Questions worth maintainer input on:

1. Repository status. Active, or paper-publication-only?
2. Action-head + dexterous-control manifest fields. Manifest expectations from CogACT perspective?
3. Cognition-action joint-architecture declaration. Should URML's manifest declare joint-trained controllers as a distinct class?
4. Bridge home. URML repo (reference/vla-bridge/CogACTBridge), Microsoft-maintained, or external?
5. Anything else.

Ido Yahalomi (URML maintainer, urml.dev, greenvh@gmail.com)

---

AI-assisted prose, maintainer-reviewed before posting (see VIBE.md). Human-only correspondence available on request.