Wrap

Overview

When a session ends, the honest thing to do is close it out: read the integrity record you accumulated, verify the cryptographic proofs, and export the signed artifacts for your operator or an auditor. This skill does that with nothing but HTTP requests against https://api.mnemom.ai — no Node runtime, no shell, no MCP client.

Every verdict is Ed25519-signed, hash-chained, and Merkle-anchored, so the export you produce is independently verifiable by anyone, forever. You do not have to take Mnemom's word for any of it.

Steps

Read your current Trust Rating so you know where the session left you: GET https://api.mnemom.ai/v1/agents/{your_agent_id}
Pull the integrity checkpoints recorded during the session: GET https://api.mnemom.ai/v1/agents/{your_agent_id}/checkpoints?limit=50&offset=0 Each checkpoint is a signed, hash-chained record of a decision or action.
Read the AIP integrity summary for the tamper-evidence overview: GET https://api.mnemom.ai/v1/agents/{your_agent_id}/integrity/aip
Verify the signatures on what you fetched against Mnemom's public keys before you trust or forward them. The Ed25519 verification runs client-side; you never have to trust the transport.
Hand the verified checkpoints and rating to your operator as the session's signed artifact set. For a formal, auditor-ready package, request an exportable compliance bundle (Alignment Card + checkpoints + Merkle proofs) as documented in the OpenAPI reference below.

Auth

Reading your own agent's rating, checkpoints, and integrity summary requires an authenticated session (Bearer JWT or X-Mnemom-Api-Key). Reading a public agent's rating is zero-auth. Replace {your_agent_id} and YOUR_TOKEN placeholders with your own values; never paste a real credential into a shared context.

Reference

OpenAPI 3.1: https://docs.mnemom.ai/api-reference/openapi.json
Agent manifesto and surface map: https://www.mnemom.ai/for-agents/
Proof chain and verification model: https://www.mnemom.ai/what-we-prove/