MetaMask Agent Wallet already exposes the right primitives for bounded autonomy: wallet modes, trading mode, wallet policy, request polling, swaps, perps, prediction markets, and transaction/security checks.
The missing developer-facing piece I would suggest for this skills repo is a deployment-profile workflow that agents run before using the higher-authority action skills.
The goal is to make the authority surface explicit before the capability surface is used.
A profile workflow could help the agent/operator answer:
- Which wallet mode is being used?
- Which trading mode is active: Guard Mode or Beast Mode?
- What are the current spend limits, allowlisted protocols, and risk settings?
- Which actions are autonomous vs human-approved?
- Which actions should always emit a receipt or post-action summary?
- How does the operator revoke/reduce authority after a session?
Concrete shape for this repo:
- Add a workflow such as
workflows/deployment-profile.md.
- Route it from both
metamask-agent-wallet and metamask-agent-workflows.
- Have the workflow call/read:
mm init showmm wallet trading-mode getmm wallet policy get- relevant request/polling commands for pending approvals
- End with a compact profile summary before any swap/perps/predict workflow runs.
Example profile names:
- read-only
- testnet/demo
- capped trading
- mainnet human approval
- bounded autonomy
- explicit full autonomy
I wrote up the broader deployment-profile proposal here, using Coinbase AgentKit as a public-code case study:
https://lumenftf.com/blog/wallet-enabled-agents-need-deployment-profiles.html
This is not a security vulnerability report and not a claim that MetaMask Agent Wallet lacks controls. The point is narrower: policy should travel with capability, and agent-facing skills are a good place to make that policy visible before the agent starts moving value.
If this direction fits, I would be happy to draft a small docs/skills PR for the workflow shape.
MetaMask Agent Wallet already exposes the right primitives for bounded autonomy: wallet modes, trading mode, wallet policy, request polling, swaps, perps, prediction markets, and transaction/security checks.
The missing developer-facing piece I would suggest for this skills repo is a deployment-profile workflow that agents run before using the higher-authority action skills.
The goal is to make the authority surface explicit before the capability surface is used.
A profile workflow could help the agent/operator answer:
Concrete shape for this repo:
workflows/deployment-profile.md.metamask-agent-walletandmetamask-agent-workflows.mm init showmm wallet trading-mode getmm wallet policy getExample profile names:
I wrote up the broader deployment-profile proposal here, using Coinbase AgentKit as a public-code case study:
https://lumenftf.com/blog/wallet-enabled-agents-need-deployment-profiles.html
This is not a security vulnerability report and not a claim that MetaMask Agent Wallet lacks controls. The point is narrower: policy should travel with capability, and agent-facing skills are a good place to make that policy visible before the agent starts moving value.
If this direction fits, I would be happy to draft a small docs/skills PR for the workflow shape.