rtime-assistant

Use when auditing, reviewing, or planning Obsidian/Zotero citation workflows for the rtime `brain` library, especially citekey crosswalks, BibTeX coverage, Zotero URI clues, Obsidian wikilinks, DocPack citation anchors, or Mac/orangepi read-only citation review.

Use when auditing, selecting samples, building, validating, or reviewing DocPack outputs for the rtime `brain` knowledge library, especially learning materials, PDFs, Office files, citation mapping, Obsidian-readable Markdown, or Mac/orangepi DocPack workflows.

Use when auditing, indexing, reviewing, or planning display modules for the rtime `brain` library, especially Obsidian vault signals, Zotero/Better BibTeX links, DocPack citation readiness, SQLite/BM25 index readiness, or Mac/orangepi read-only library workflows.

Use when configuring, auditing, or debugging rtime-assistant agent tooling, MCP client snippets, tool installation status, validation plans, context-lane routing, or read-only runtime snapshots for AI-agent control surfaces.

Use when inspecting or troubleshooting the rtime-assistant runtime, Feishu/Lark bridge candidate, run logs, queue/access/output modules, Docker production Compose checks, systemd templates, or Mac/orangepi runtime diagnostics without changing live services.

Use when inspecting, planning, or reviewing rtime assistant reminders, schedulers, notifications, workflow automation, timer health, or automation run-log surfaces without sending messages or mutating runtime state.

Use when planning dynamic context unlocking, building ContextUnlockPlan or Context Pack skeletons, explaining why context should be loaded, routing requests across brain/rtime-hub/runtime evidence, or designing why-context behavior without reading sensitive data.

Use when reading, auditing, or planning project status, device status, contact-directory panels, service/task/deployment context, or Mac/orangepi rtime-hub workflows without modifying the rtime-hub fact store.

Use when an assistant needs one gated entry point to the rtime brain library and assistant data — searching/indexing/auditing the brain, hub, profile, review, automation, runtime read surfaces, or making one of the three narrow settings writes (context sources, memory candidates, reminders) — behind a single central permission gate (personal-data gating and output redaction are policy switches, both OFF in the current single-owner deployment), and metadata-only audit.

Use when reviewing, planning, or packaging rtime assistant profile and policy adjustments, especially persona, prompt layering, model policy, permission policy, output style, memory behavior, sensitive-data rules, Codex/Claude skills, plugins, or MCP exposure on Mac and orangepi.

Use when registering, listing, cancelling, testing, or diagnosing rtime Feishu reminders and scheduled reminder delivery. Trigger for Chinese or English reminder requests such as "提醒我", "定时提醒", "闹钟", "remind me", or "schedule a reminder"; use the rtime JSONL reminder path and never Claude Code Cron for phone push reminders.

Use when building, inspecting, or planning review-console surfaces for rtime-assistant, especially memory candidates, failed runs, run-log summaries, permission audit, standards audit archives, tooling health, context-plan review, or future Web console governance panels.

Use when the user asks to file, organize, ingest, upload, or batch-process PDFs, courseware, images, web clips, Feishu attachments, Obsidian attachments, or any material into the rtime brain library.

登记、查看、取消 rtime 飞书定时提醒。当用户说“提醒我”“定时提醒”“闹钟”“X分钟后叫我”时使用；必须写入 rtime reminders JSONL，不使用 Claude Code Cron。

Use when registering, listing, cancelling, testing, or diagnosing rtime Feishu reminders and scheduled reminder delivery. Trigger for Chinese or English reminder requests such as "提醒我", "定时提醒", "闹钟", "remind me", or "schedule a reminder"; use the rtime JSONL reminder path and never Claude Code Cron for phone push reminders.