Extract durable facts from transcripts into a Memory graph via a configured AI provider.
npx skills add https://github.com/reddb-io/red-skills --skill extract-reddb-io
Please help me install this Agent Skill. Skill: extract Source: https://github.com/reddb-io/red-skills/tree/main/plugins/memory/skills/core/extract Command: npx skills add https://github.com/reddb-io/red-skills --skill extract-reddb-io
INFERRED nodes and edges into the Memory graph, reporting how many facts and edges were created. - Verification via Recall: Confirms the new facts are discoverable by running a targeted recall query after extraction. - Use Case: After a long debugging session, pipe the session transcript into the extract command so the root cause and the fix constraint are preserved as graph facts that future sessions can recall. ## Quick Start Extract the durable decisions and gotchas from my session transcript file debug-session.log into the Memory graph, then verify them with a recall query.High-intent search queries and answers about installing and using this skill.
Run the extract command with a transcript file or pipe it via stdin using the plugin's bootstrap script. The configured AI provider identifies durable facts and upserts them as INFERRED nodes and edges into the Memory graph.
Extract durable operational learning: decisions, root causes, gotchas, why-notes, and failed attempts that explain constraints. Avoid secrets, credentials, personal data, PR numbers, commit SHAs, and transient task-completion logs.
No. Extraction requires graph mode and a provider block in the plugins.memory section of .red/config.yaml. If either is missing, the skill stops and explains rather than falling back to markdown notes.
Use extract for rich transcripts containing multiple decisions and reasoning traces. Use the store command for a single hand-authored fact. Extraction mutates the graph, so do not run it just to summarize a conversation.
Run a targeted recall query for one or two extracted topics after extraction. The recall command confirms whether the new INFERRED facts are discoverable in the graph.
marketiv-idReports actual session token usage and savings from the Claude Code session log.
Nerfherder16Store explicit memories with domain, tags, and classification via a memory-store endpoint.
ayo-oooStore and retrieve persistent memories as Markdown files with TOML frontmatter across sessions.
onebrain-aiAudit and prune persisted memory entries by reviewing MEMORY-INDEX.md files.