Ledger’s Agent Intent queue lets background agents (Grok Bot, Muse, Hermes, OpenClaw, Dots) prepare actions—e.g. Grok Bot watches an inbox and drafts an invoice payment—while you review asynchronously and physically confirm on a Ledger signer Secure Screen. Agents never hold signing keys; coding agents can use Agent Stack CLI/skills too.
Muse Code journals sessions to an append-only event log. Author a focused golden fixture (records + expected_model_context), then `muse trace inspect --fixture` replays projection with no model, tools, or network. Gate merges on zero diffs via replay_gate.sh / GitHub Actions—catch harness drift before it ships.
Viaknox’s open-source Tune-Up (v0.3.4, MIT) installs as a Grok Bot template/plugin, reviews bots you put in scope against five checks, turns misses into tests, and pastes small instruction fixes only after you approve. Skill index loads the right SKILL.md only when needed (claimed ~77% shorter always-on instructions). Free connector also works with Claude/Cursor; ChatGPT beta.