Stats
Actions
Tags
From agentmemory
Deletes specific observations or sessions from agentmemory via API after user confirmation. Use for privacy requests like 'forget this', 'delete memory', or targeted data removal.
How this skill is triggered — by the user, by Claude, or both
Slash command
/agentmemory:forgetThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
The user wants to remove data from agentmemory: $ARGUMENTS
The user wants to remove data from agentmemory: $ARGUMENTS
IMPORTANT: This is a destructive operation. Always confirm with the user before deleting.
Steps:
First, search for matching observations:
curl -s -H "Content-Type: application/json" \
-H "Authorization: Bearer ${AGENTMEMORY_SECRET:-}" \
-X POST "http://${AGENTMEMORY_URL:-localhost:3111}/agentmemory/search" \
-d '{"query": "<SEARCH_TERM>", "limit": 20}'
Show the user what was found and ask for confirmation
If confirmed, delete via:
curl -s -H "Content-Type: application/json" \
-H "Authorization: Bearer ${AGENTMEMORY_SECRET:-}" \
-X POST "http://${AGENTMEMORY_URL:-localhost:3111}/agentmemory/forget" \
-d '{"sessionId": "<ID>"}' # or {"observationIds": ["id1", "id2"]}
Confirm deletion to the user
Never delete without explicit user confirmation.
npx claudepluginhub rohitg00/agentmemory --plugin agentmemorySelectively and reversibly forget stored memories/concepts by query or date, soft-deactivating them from retrieval without deletion. Useful for privacy or decluttering the kongcode graph.
Supersedes memory entries matching a query in CLAUDE.local.md, registers/*.md, and daily/*.md. Preserves history without deletion, annotates as outdated, updates metadata.json.
Audits and selectively forgets stored Claude Code memories. Use when memory is large/uncurated, project state has shifted, or retrieval quality degraded.