Bidirectional sync between gbrain entity pages and the Karpathy Markdown Knowledge Base with v0.10 convention enforcement
Generate entity-aware morning briefing from gbrain — active deals, recent meetings, people in play, and upcoming timelines. Posts a structured Slack thread to briefing", "entity briefing", "who's in play", "active deals", "gbrain morning", "gbrain-briefing", "엔티티 브리핑", "gbrain 아침", "인맥 브리핑", "활성 딜", or wants an entity-centric morning overview. Do NOT use for daily stock analysis (use today). Do NOT use for general web search (use parallel-web-search). Do NOT use for Cognee knowledge graph queries (use cognee).
Wrapper around gbrain's ingest capabilities adapted to the project's file-first pipeline conventions. Accepts pipeline output files (`outputs/` manifests), extracts entity-relevant content, and routes it to gbrain via MCP tools. Supports `--ambient` mode for lightweight signal detection. Enforces gbrain v0.10 conventions: brain-first lookup, Iron Law of Back-Linking, inline citations, and subject-based filing.
Comprehensive brain health maintenance: multi-dimension diagnostics, benchmark scoring, heartbeat loop, compiled-truth refresh, and Slack report output. Wraps `gbrain doctor --json`, `gbrain health` (text dashboard), `gbrain stats`, and `gbrain search` into an actionable maintenance workflow.
Meta-orchestrator that decomposes user goals into agent execution plans and (optionally) pursues them across turns via Goal Mode. Maintains a persistent wiki-style TODO backlog under state/todos/. Searches available skills, classifies task complexity, selects optimal agent composition, produces a numbered dispatch plan with model routing and dependency ordering, and persists multi-turn objectives with pursuing/paused/achieved/unmet/budget-limited lifecycle states. Use when the user asks to "plan how to do X", "jarvis", "decompose this goal", "자비스", "goal mode", "pursue this until done", "jarvis goal", "jarvis plan", "자비스 계획", "자비스 목표", "long-running objective", "multi-turn goal", "자비스 TODO", "jarvis todo", "보고서 작성해줘", "리포트 만들어줘", "보고서 생성", "분석 보고서", "페르소나 분석", "관점 분석해줘", "딥 리서치", "리서치해서 보고서로", or wants intelligent skill composition planning with optional persistent goal pursuit. Pipeline Mode auto-triggers for registered pipelines (report, persona, deep-research, research-report) and executes immediately with
Builds a personalized strategic + tactical asset-allocation report from a risk profile and the current macro regime, sized in KRW with concrete instruments (KR-listed ETFs preferred), rebalancing bands, DCA schedule, and risk guardrails. Use when the user asks "내 돈 어떻게 분산/배분해?", "자산배분 해줘", "포트폴리오 구성", "1억 어디에 투자", "asset allocation", "how to diversify my money", "portfolio construction". Do NOT use for single-stock pre-earnings analysis (use equity-research-agent), daily trade entries (use daily-trade-plan), or reviewing an existing brokerage account's holdings (use toss-portfolio-recon / portfolio-report-generator).
Generate session-specific trade plans: KR session (morning, before Korean market open) or US session (evening KST, before US market open). Each has its own market context, catalysts, playbook, and position sizing. Use when "오늘 매매 전략", "한국장 전략", "미장 전략", "US session plan", "daily trade plan", "미국장 전 전략", "tonight's trading strategy". Do NOT use for backtested mechanical strategy cards (use daily-strategy-engine). Do NOT use to place real orders (use tossinvest-trading). Do NOT use for FX conversion (no safe tool — human-in-the-loop only).
Autonomous daily intelligence pipeline that cross-references HF papers, models, datasets, spaces, leaderboards, and community activity to detect emerging AI trends before they go mainstream. Supports optional topic filtering (LLM, multi-LLM, video-generation, etc.) for focused scans. Produces a scored intelligence report and distributes to Slack + Notion. Use when the user asks to "run daily AI radar", "trending intelligence", "AI research radar", "detect AI trends", "daily paper intelligence", "AI 트렌드 레이더", "일일 AI 인텔리전스", "트렌딩 분석", "연구 레이더", "HF 종합 트렌딩", or wants to know what's trending across the HF ecosystem today. Do NOT use for full paper review with PM analysis (use paper-review). Do NOT use for daily stock analysis (use today). Do NOT use for general web research only (use parallel-web-search). Do NOT use for paper browsing without cross-referencing (use hf-papers). Do NOT use for topic-only scan without full cross-referencing (use hf-topic-radar). Do NOT use for leaderboard-only tracking (use hf-leade