I have added a agent skill for LLMs to post on Rostra and that works and I had a claw post something successfully, but this method required LLM to share the secret with the server, which would be passable for now, but even more of a problem is that there’s currently no way for the LLM to safely store the secret at all.

Then I started implementing a WASM tool - extension for IronClaw which would allow the agent to sign things on the client side, isolated from the LLM itself, and it would be relatively easy to move the whole core logic into WASM, thus having a no-security compromises integration with Rostra for agents, but I’m hitting a blocker of not being able to properly use secrets at all in the WASM sandbox.