Audits and repairs knowledge-base health across stale pages, backlinks, citations, and embeddings.
npx skills add https://github.com/AMC-JTC/gbrain-1 --skill maintain-amc-jtc
Please help me install this Agent Skill. Skill: maintain Source: https://github.com/AMC-JTC/gbrain-1/tree/main/plugin/skills/maintain Command: npx skills add https://github.com/AMC-JTC/gbrain-1 --skill maintain-amc-jtc
gbrain doctor --remediate with a target score and cost cap to automatically walk a dependency-ordered repair plan, or performs manual per-dimension fixes when judgment is needed. - Dream cycle maintenance: Consolidates conversation transcripts into long-term memory via triage-gated synthesis and pattern detection, with idempotency, privacy filters, and spend cooldowns. - Use Case: Ask for a weekly brain health check and receive a structured report table showing issues found, fixed, and remaining per dimension, plus benchmark results verifying search quality has not regressed. ## Quick Start Ask the assistant to run a full brain health check and fix any stale pages, broken backlinks, and missing citations it finds.High-intent search queries and answers about installing and using this skill.
Run gbrain doctor --json or gbrain health to get a dashboard of all dimensions including stale pages, orphans, dead links, RLS status, and schema version. For automatic repair, use gbrain doctor --remediate --yes --target-score 90 with a cost cap.
Review orphan pages to decide whether they need inbound links from related pages or deletion. For backlinks, check that every entity mention has a corresponding back-link on the entity page, and add missing ones in the Timeline or See Also section.
gbrain dream runs the full maintenance cycle: lint, backlinks, sync, synthesize, extract, patterns, embed, and orphans. The synthesize phase triages conversation transcripts with a cheap judge model, then fans out synthesis subagents that write reflections, ideas, and timeline entries.
Yes. dream.synthesize.cooldown_hours (default 12) caps run frequency, dream.triage.max_ms bounds judging time, and gbrain doctor --remediate accepts --max-usd as a hard cost cap that refuses submission when the plan would exceed it.
Run gbrain extract links and gbrain extract timeline after major imports or bulk content edits. New gbrain put calls auto-create links via the auto-link hook, so extraction is mostly a one-time backfill; verify with gbrain stats and gbrain health coverage percentages.
Some dimensions have hard ceilings: an empty brain with no entity pages caps graph_coverage at 70, and an unconfigured embedding key caps at 60. The remediate command bails with a list of what is missing rather than looping indefinitely.
paxlabs-incIngest raw sources into an interlinked markdown knowledge base with provenance tracking.
hck717Extract financial metrics and map relationships from company filings.
weipingEnrich newly created person pages with structured context for downstream question answering.
IgorGanapolskyValidates local graph-based wiki navigation retrieval against vector-only search baselines.