메뉴
Import external sources into the brain — articles, videos, papers, web pages, handover documents. Use when the user asks to ingest, import, or add external content to the knowledge base, or run /brain-ingest.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Update Fritz Local to the latest version. Pulls from git, symlinks new skills and managed hooks, runs pending migrations, and reports changes. Use when the session-start hook reports an update is available, or run /update manually.
Create a structured handover document for continuing work in a fresh agent session. Preserves knowledge before handing over: compiles pending captures, ingests session decisions/patterns, syncs if configured, then writes the handover document. Use when the user asks to hand over, hand off, create a handover, wrap up for continuation, save session state, prepare a fresh session, or run /handover.
Interactive runbook to provision or reconfigure the Fritz Local Brain Docker service. Asks the full question set (LLM backend, embeddings, scheduler, autostart, context injection, optional one-time backlog drain, API token), confirms a summary, then delegates ALL execution to the PROV1 provisioning engine (scripts/local-brain-service.py provision). Does not write any config or start any container before every answer is confirmed. Use when the user wants to set up, provision, or reconfigure the Local Brain Docker service, or run /brain-service-setup.
Set up the Fritz Local brain overlay for a new vault. Explores directory structure, presents findings, asks questions interactively, then creates structure based on human answers. Handles per-project directories, common/ shared area, index files, .fritz-local.json creation, and context injection configuration. Use when the user asks to set up a brain vault, add a project, or run /brain-setup.
Push knowledge articles from brain vaults to external systems (AFFiNE, Obsidian, etc.). Brain markdown is source of truth — external systems are views. Use when the user asks to sync, push, or export brain knowledge, or run /brain-sync.
Search and synthesize knowledge across all brain vaults. Use when the user asks to search the brain/knowledge base, find what they know about a topic, recall a decision/pattern/lesson, or run /brain-query.
| name | brain-ingest |
| description | Import external sources into the brain — articles, videos, papers, web pages, handover documents. Use when the user asks to ingest, import, or add external content to the knowledge base, or run /brain-ingest. |
Import external sources into the brain — articles, videos, papers, web pages. This is Karpathy's "raw → wiki" ingest operation.
Activate when the user asks to:
/brain-ingestDetermine the source type and obtain the content:
Write the raw content to ~/.brain/capture/inbox/ as a provenance-tagged
capture. Filename: YYYY-MM-DD-<source-slug>.md. Include frontmatter:
---
type: capture
title: "<source title>"
sources:
- <original URL or path>
created: <today>
agent_last_edit: <agent>
---
For content mirrored from an external system (e.g. via the service's mirror agent), the frontmatter also carries provenance fields written by the mirror pipeline:
---
type: capture
title: "<entry title>"
source: "<target-name> (<target-kind>)"
mirrored_at: "<ISO-8601 timestamp>"
mode: "full" | "index-only"
pointer: "<target-name>:<relpath-or-id>"
---
Do not write these provenance fields manually when ingesting from a URL or file — they are reserved for the service mirror path.
After writing the inbox capture, call the configured Fritz Local processing path:
settings.local_brain_service.enabled: true and auto_compile_on_ingest is not false, trigger the Local Brain compile path (service first, local fallback) so the capture does not silently pile up.~/.brain/capture/inbox/ and report that compile is pending/manual.If live processing is unavailable and manual compile is needed, read the raw source and create a compiled knowledge article. In registry-free mode, write to ~/.brain/knowledge/<scope>/<section>/<slug>.md (scope = common or project slug; section = decisions, lessons, runbooks, or context). In vault mode, write to the appropriate knowledge/ subfolder. Either way:
sources pointing to the inbox file.brain/log.md:YYYY-MM-DD HH:MM | INGEST | <agent> | Ingested "<title>" from <source type>, created <article path>
User: "Ingest this video: https://youtu.be/1FiER-40zng"
capture/inbox/2026-04-07-cole-medin-second-brain.mdknowledge/300_Resources/ai-second-brain-architecture.md with structured summarycapture/inbox/ — it's the provenance chainrelated frontmatter