protect-mcp-setup
Configure Cedar policy enforcement and Ed25519 signed receipts
by wshobsonFree
Description
protect-mcp-setup is an agent skill from the wshobson/agents collection for configuring policy enforcement and signed audit receipts for Claude Code tool calls. Claude Code can run powerful tools such as Bash, Edit, Write and WebFetch, and by default there is no record of what was run and no check against a policy beforehand. The skill sets up the protect-mcp approach to close those gaps. Each tool call is first evaluated against a Cedar policy, the open authorization engine from AWS, and a Cedar deny is authoritative, so a blocked call does not run. Each call that runs produces an Ed25519-signed receipt that records the name of the tool that ran, signed with your key, and the receipts can be checked offline with the verification tool published on npm, without a server or an account. The skill is for projects that need cryptographic audit trails of agent actions and a policy that is enforced before execution rather than reviewed afterwards. Setup involves choosing the policies and managing the signing key, and the signing key should be kept secret and backed up, because receipts can only be verified against the key that signed them. The skill is installed with the skills CLI from the wshobson/agents repository. Its skills.sh page showed about 5.8K installs and passing results from three security audits at the time of the scan.…
Highlights
- Cedar policy evaluation before each tool call
- Ed25519-signed receipts for each tool run
- Offline receipt verification
- Audit trail for Claude Code tools
- Setup for policies and signing keys
- Installs with the skills CLI
Ratings & Reviews
No ratings yet
Be the first to rate protect-mcp-setup.
People also visit
review-agent-setup
Configure human-in-the-loop gating for AI agent review actions in Claude Code
protect-mcp-setup FAQ
What does this skill configure?
Cedar policy enforcement and Ed25519-signed receipts for Claude Code tool calls.
What happens when a policy denies a call?
The Cedar deny is authoritative, so the call is not run.
How are receipts checked?
Offline, with the verification tool published on npm, without a server or an account.
What do I need to keep safe?
The signing key. Receipts can only be verified against the key that signed them, so keep it secret and backed up.
How do I install it?
Run npx skills add https://github.com/wshobson/agents --skill protect-mcp-setup.
What security audits does it have?
skills.sh lists three audits: Gen Agent Trust Hub Pass, Socket Pass and Snyk Pass.
What are the best protect-mcp-setup alternatives?
Popular alternatives to protect-mcp-setup include signed-audit-trails-recipe, block-no-verify-hook and review-agent-setup. See all protect-mcp-setup alternatives