Who it is forA security-minded engineer or reviewer who needs evidence of every memory change in a DeepSeek Harness profile.
The problemWithout a record, it is hard to tell which memory entries came from the model and which were approved by a person.
The outcomeA list of memory writes with their approval outcomes, plus a visible note where the session log does not record memory events.
What you need first
- Node ^22.19.0 or >=24.0.0
- DeepSeek Harness with the dsh CLI installed
- A dsh profile, such as the web profile used in the examples
Steps
- Run
/memory auditto list the plugin audit table. - Read the gap line if it appears, which means the session log does not record memory event types.
- Match each audit row with its approval/asked and approval/decided pair in the session log.
- Set auditRetentionDays in the memento config if you need audit rows removed after a set number of days; 0 keeps them forever. (inferred, check the README)
Prompt for your agent
Using the dsh-memento repo on profile <profile>, run /memory audit and list each write with its outcome. Expected output: rows from the plugin audit table, plus the session-log gap line if it is shown. Verify that each write has a matching approval/asked and approval/decided pair in the session log.
Caveat. The README says the session log does not yet know the memory event types, so the audit is only complete when the approval pair and plugin table are both checked.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for PerryLink/dsh-memento and "Audit what the model saw and what it wrote"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer