Reference
Catalogs
The shipped MCP servers, agents, skills, and routines in four generated summary tables — each row linking into the in-depth infrastructure pages.
Four catalogs cover everything agentic that ships with Command Center: 17 MCP servers exposing 290 tools, 48 agent definitions, 23 skills, and 24 routine definitions. Each table is generated straight from the shipped definitions on your disk — the same files your agents load — so the names and counts on this page cannot drift from the product. For the narrative behind each catalog, start at Agentic infrastructure.
MCP servers
Each server is a local process your agent sessions load from the install
root's .mcp.json manifest. Together they are the tool fabric: every
capability an agent reaches — memory, the knowledge graph, email, harvesting,
video, terminals, windows, missions — goes through one of these servers. The
in-depth, tool-by-tool reference lives under
MCP servers.
| Server | Tools | Purpose |
|---|---|---|
shared-agent-memory | 7 | Cross-agent shared memory — durable findings, decisions, and patterns with hybrid keyword and semantic search. |
knowledge-base | 22 | Document ingest and workspace lifecycle for the knowledge graph, plus semantic search on full-depth workspaces. |
gemini-direct | 1 | Direct Gemini queries — a single query tool with per-call model override. |
arcadedb | 12 | The graph arm of the knowledge base — direct queries, traversal, discovery, and node/edge editing over the unified knowledge graph. |
observer | 23 | News harvesting — manages sources, triggers and monitors harvests, and lets agents browse, search, and curate the item store. |
email | 19 | Email operations over standard IMAP/SMTP — read, search, send, reply, forward, and a shared queue for automated triage. |
imagegen | 8 | Image generation through the local image service — single images, batches, section illustrations, and generation history. |
x-mcp | 22 | The X (Twitter) drafting surface — drafts, pre-flight compliance and cost checks, and analytics; publishing always stays human-approved. |
products-mcp | 26 | The products pipeline — create products, dispatch the authoring stages, manage scope, generate covers, and export. |
hyperframes-mcp | 19 | The HTML-first video pipeline — briefs, site capture, the three authoring stages, lint/inspect, templates, and renders. |
remotion-mcp | 25 | The programmatic video pipeline — briefs and revisions, the composition catalog and schemas, media probes, and renders. |
x-grok | 6 | Grok access through your own subscription — chat, X search, image generation, speech, and video. |
footage-mcp | 15 | Raw-footage preparation — sources, transcription, proposed cuts, trims and splits, grades, and clean-cut renders. |
calendar-mcp | 14 | calendar-mcp Server — agent-facing calendar + scheduling capability surface. |
cc-terminals | 31 | The terminal control plane — open, drive, read, and close Command Center terminal panes and workspaces, including mission-fleet dispatch. |
cc-window-control | 28 | Window and view control of the running app — window management, view navigation, and knowledge-graph canvas control. |
missions | 12 | The mission registry — missions, operations, and projects with transitions, events, documents, and lint. |
Agents
Agent definitions live at .claude/agents/ under your install root — plain
Markdown files you can read and edit. The roster mixes roles you dispatch
yourself (researchers, auditors, developers) with pipeline roles the
background services dispatch for you (the news desk, the products pipeline,
the video pipelines). The in-depth roster is at
Agents.
| Agent | Purpose |
|---|---|
breaking-writer | A specialist agent that extracts time-sensitive breaking-news signals from recent observer reports and writes one Live Intel Feed ticker item per distinct signal. |
context-architect | Context engineering specialist. Assesses a task or situation, queries all memory layers (shared-agent-memory, knowledge-base, documents), identifies relevant files and knowledge, and composes a structured context artifa… |
critique-validator | State-of-the-art critique and validation agent for evaluating any agent output — research reports, implementation plans, architectural decisions, proposed solutions. |
deep-researcher | Multi-phase deep research agent for comprehensive investigation of any topic. |
deepdive-author | Produces a verbose, primary-source capabilities / architecture / infrastructure / order-of-operations DEEP-DIVE document on any subject — a feature, a subsystem, a whole codebase, or an external repo — purpose-built to… |
dependency-checker | Dependency audit specialist. Use when checking for outdated packages, security vulnerabilities in dependencies, license compliance, or dependency conflicts in any project. |
doc-organizer | MD report lifecycle and organization specialist. Use when organizing, consolidating, archiving, or auditing the MD file ecosystem under documents/general_reports (and any other per-component report hubs the project defi… |
feature-auditor | Senior systems auditor for deep feature audits. Use when you need a comprehensive production-readiness, correctness, and architecture audit of any feature in the platform. |
feature-developer | Full-stack feature implementation specialist. Use when adding new functionality to any project — new backend routes, services, frontend components, state stores, database models, migrations, or any combination. |
footage-editor | Prepares ONE raw clip for use in a video project, per a natural-language operator instruction. |
hyperframes-brief-author | First agent in the 3-agent Hyperframes (HTML-first) video pipeline. |
hyperframes-html-author | Third agent in the 3-agent Hyperframes video pipeline. |
hyperframes-planner | Second agent in the 3-agent Hyperframes (HTML-first) video pipeline (after hyperframes-brief-author). |
instagram-content-creator | A specialist agent that drafts Instagram posts (caption + image prompt chain) for the brand's Creator-account manual-publish workflow. |
kg-ingester | Claude-native knowledge-graph ingest specialist. Use when you need to ingest one or more documents into a regular LightRAG workspace WITHOUT calling Gemini for entity extraction — reads the file(s), extracts entities an… |
mission-commander | Operating role for a mission commander — the grounded terminal session that drives one mission's fleet to its Success Conditions. |
mission-worker | Operating posture for a mission worker — a grounded terminal that executes ONE operation for its mission commander. |
news-illustrator | A specialist agent that reads a single intelligence article in full and composes a rich, visually-specific hero-image prompt informed by the article's actual content — not just its title + excerpt. |
news-scout | A specialist agent that reads a curated pool of observer-report files and proposes a ranked set of briefing-worthy story clusters for the `news-writer` agent to synthesize. |
news-translator | A specialist agent that produces native-sounding professional translations of source intelligence articles into the product's configured target locales, published on the platform's website. |
news-writer | A specialist agent that transforms curated observer data into publication-ready intelligence content for the platform. |
newsletter-writer | A specialist agent that curates a bilingual email digest from recent briefings + breaking news + (optionally) a featured roundup or report. |
observer-agent | An AI news intelligence analyst for the platform's multi-source harvesting infrastructure. |
product-auditor | Stage 2.C Phase 3 + Stage-3 on-demand — the product auditor/director (formerly product-architect, repurposed 2026-05-28 under Fork A). |
product-composer | Given the product's sighted material + working thesis, proposes a foundation outline for an in-depth digital product. |
product-consolidator | Synthesis pass for a single section. Reads N existing research-note .md files (operator-selected via CC checkboxes) and produces ONE consolidated, deduplicated artefact preserving every unique insight while merging over… |
product-draft | The product-draft agent, a long-form writer producing one section of an in-depth digital product (playbook, blueprint, framework, audit) for the platform's `products/` catalog. |
product-human-author | Stage 3 Sub-flow 3.H — human-facing assembly (PER-SECTION). |
product-illustrator | A specialist agent that reads a product's structural and substantive artifacts and composes a rich, visually-specific cover-image prompt for an editorial-tier digital product. |
product-research-scout | Planning step of a research dispatch. Reads the topic_index map + coverage state + dispatch config and produces a per-dispatch manifest of (doc_path, line_range, why) tuples for the downstream product-researcher to dril… |
product-researcher | Focused reading + EXTRACTION step of a research dispatch (Stage 2.B; also reused by Stage-3 scope_research). |
product-scope-refiner | Stage 3 Sub-flow D Phase 3 — section refiner (TRANSFORM step, re-chartered 2026-06-01). |
product-scout | A specialist agent that researched and structured the foundation for an in-depth digital product (playbook, blueprint, framework, or audit) destined for the `products/` catalog. |
product-section-finalizer | Stage 3 Sub-flow 3.G — section finalizer. Takes ONE section's latest content (the refiner's processed version when present, else the sanitised scope copy) + the product goal + optional operator instructions + optional r… |
product-sighter | Reads every un-sighted-or-stale seed file for a single product and emits ONE JSON dispatch manifest listing per-doc structured topics (tag + display + line_range + anchor + summary) plus a 1-2 sentence doc_summary per d… |
professor-gemini | Google Cloud / Vertex AI / Gemini / Veo domain expert and consultant. |
quick-researcher | Lightweight research agent for focused investigations that don't need deep-researcher's exhaustive multi-phase pipeline. |
refactorer | Code refactoring specialist for improving code quality without changing behavior. |
report-scout | A specialist agent that reads a curated pool of intelligence source files and proposes a ranked set of report-worthy topic clusters for the `report-writer` agent to synthesize. |
report-writer | A specialist agent that produces publication-ready deep-dive intelligence reports for the website. |
roundup-writer | A specialist agent that produces ONE weekly multi-topic AI intelligence roundup per invocation, summarizing the week's themes from briefings + observer reports + breaking items. |
schedule-executor | Executes ONE scheduled calendar entry — the agent behind a Command Center Calendar event. |
support-email | Automated support email handler. Processes the email queue — reads new emails, classifies them, queries KB for context, drafts responses with HITL approval, sends replies, and logs interactions to shared-agent-memory. |
video-brief-author | First-pass video pipeline agent — reads the operator's freeform video brief, knows the high-level categories of capabilities the system can produce (text effects, comparisons, code walkthroughs, particle backgrounds, br… |
video-composer | Agentic two-phase video composition specialist. Reads a video brief (operator's prompt + assets + hints) along with the Remotion composition catalog and capability docs, then produces a complete render proposal — both m… |
video-script-author | Second-pass video pipeline agent — takes an enriched brief (with hints.creative populated by video-brief-author or the operator) and produces a concrete scene-by-scene production script. |
x-content-creator | A specialist agent that drafts X / Twitter posts (single tweets, threads, replies, polls) from curated source artefacts for the brand's full-publish X workflow. |
yt-video-analyzer | Analyzes harvested YouTube video items end-to-end from their multi-medial raw artifacts (transcript + extracted keyframes + companion metadata). |
Skills
Skills are packaged expertise an agent loads on demand — repeatable methods
with their reference material bundled next to them. They live at
.claude/skills/ under your install root, one folder per skill. The in-depth
catalog is at Skills.
| Skill | Purpose |
|---|---|
animejs | Anime.js adapter patterns for HyperFrames. Use when writing Anime.js animations or timelines inside HyperFrames compositions, registering animations on window.__hfAnime, making Anime.js seek-driven and deterministic, or… |
codebase-viz | Generate an interactive HTML visualization of the codebase structure. |
command-center-design | Use this skill to generate well-branded interfaces and assets for Command Center, either for production or throwaway prototypes/mocks/etc. Contains essential design guidelines, colors, type, fonts, assets, and UI kit co… |
commit | Create a well-structured git commit with conventional commit format. |
contribute-catalog | Author a new HyperFrames registry block (caption style, VFX block, transition, lower third) or component (text effect, overlay, snippet) and ship it as an upstream PR to the hyperframes repo. |
css-animations | CSS animation adapter patterns for HyperFrames. Use when authoring CSS keyframes, animation-delay based timing, animation-fill-mode, animation-play-state, or CSS-only motion that HyperFrames must seek deterministically… |
deep-research | Perform deep research on a topic within this codebase. |
deepdive | Produce a verbose, primary-source capabilities / architecture / infrastructure / order-of-operations DEEP-DIVE document on any subject — a feature, a subsystem, a whole codebase, or an external repo — purpose-built to f… |
footage-prep | Edit raw recorded footage (an mp4 explainer, a talking-head take, screen capture) into a clean, condensed cut — remove fillers/retakes/dead silence, pick the best take, color-grade, burn captions, composite overlays, an… |
frontend-feature | Scaffold a new frontend feature module following the project's established patterns. |
gsap | GSAP animation reference for HyperFrames. Covers gsap.to(), from(), fromTo(), easing, stagger, defaults, timelines (gsap.timeline(), position parameter, labels, nesting, playback), and performance (transforms, will-chan… |
hyperframes | Create video compositions, animations, title cards, overlays, captions, voiceovers, audio-reactive visuals, and scene transitions in HyperFrames HTML. |
hyperframes-cli | HyperFrames CLI dev loop — `npx hyperframes` for scaffolding (init), validation (lint, inspect), preview, render, and environment troubleshooting (doctor, browser, info, upgrade). |
hyperframes-media | Asset preprocessing for HyperFrames compositions — text-to-speech narration (Kokoro), audio/video transcription (Whisper), and background removal for transparent overlays (u2net). |
hyperframes-registry | Install and wire registry blocks and components into HyperFrames compositions. |
lottie | Lottie and dotLottie adapter patterns for HyperFrames. |
remotion-to-hyperframes | Translate an existing Remotion (React-based) video composition into a HyperFrames HTML composition. |
skill-builder | Build a new Claude Code skill from scratch. Use when the user wants to create a custom skill, automate a workflow, or add a new slash command. |
tailwind | Tailwind CSS v4.2 browser-runtime patterns for HyperFrames compositions. |
three | Three.js and WebGL adapter patterns for HyperFrames. |
typegpu | TypeGPU and raw WebGPU adapter patterns for HyperFrames. |
waapi | Web Animations API adapter patterns for HyperFrames. |
website-to-hyperframes | Capture a website and create a HyperFrames video from it. |
Routines
Routine definitions are the automation catalog: each file at
.claude/routines/ describes one job the routines engine can schedule or run
on demand, executed by a fresh agent process each time. Type distinguishes
scheduled definitions from on-demand ones, Cadence is the shipped schedule,
and Default shows whether the definition ships enabled — you turn any of
them on or off from the Routines view. The in-depth catalog is at
Routines.
| Routine | Type | Cadence | Default | Purpose |
|---|---|---|---|---|
blackboard-review | recurring | weekly | on | Audit grand-blackboard files for stale references and outdated entries |
breaking-writer | recurring | 30min | on | Extract time-sensitive breaking-news items from recent observer reports (email, youtube, x-accounts). |
cortex-lint | recurring | weekly | off | Run KG health checks on knowledge-base workspaces and file the report |
doc-lifecycle | recurring | daily | on | Scan and manage document lifecycle flags in documents/general_reports/ |
email-agent | on-demand | on demand | off | Works the email queue via the multi-account email MCP, drafts under HITL as structured JSON sidecars, always writes a run report |
enrichment | on-demand | on demand | on | Extract entities, relationships, descriptions, and summaries from documents marked for enrichment |
instagram-content-creator | on-demand | on demand | on | Draft Instagram posts (caption + hero image) from intelligence reports, observer reports, and operator-supplied custom files. |
kg-artifact-cleanup | recurring | weekly | on | Reconcile the workspace registry against ArcadeDB and context artifacts — report (or prune) orphans in all three drift directions |
memory-hygiene | recurring | weekly | on | Deduplicate and consolidate shared agent-memory entries, flag stale facts |
news-illustrator | recurring | 15min | on | Auto-generate hero images for fresh non-breaking intelligence articles. |
news-translator | recurring | on demand | on | Produces native-sounding professional translations of intelligence articles into the product's configured target locales. |
news-writer | recurring | daily | off | Generate impact-ranked, source-grounded briefings per run. |
newsletter-writer | recurring | on demand | on | Curate a bilingual email digest from recent briefings + breaking news + (optional) lead reference. |
observer-report-twitter | recurring | 30min | on | Digest unread harvested items into intelligence reports (daemon pre-filters, 1 run per batch with items) |
observer-report | recurring | 30min | on | Digest unread harvested items into intelligence reports (daemon pre-filters, 1 run per batch with items) |
product-draft | on-demand | on demand | on | Stage 4 of the Products feature pipeline. Per-section sequential fan-out — one fresh `claude -p` per section running the `product-draft` agent. |
product-research | on-demand | on demand | on | Stage 2 of the Products feature pipeline. Dispatches the `product-scout` subagent (sonnet) against a curated source pool built from observer raw items + observer reports + briefings + reports + roundups + breaking + (op… |
product-sight | on-demand | on demand | on | Stage 2 Sub-flow Σ (pre-stage). Dispatches the `product-sighter` subagent (opus) against a product's seed_files + seed_folders (Q13 scope). |
report-writer | recurring | on demand | on | Generate 1-5 deep-dive intelligence reports per run. |
roundup-writer | recurring | on demand | on | Generate one weekly multi-topic AI intelligence roundup per run from pre-resolved sources (briefings + observer reports + breaking, last 7 days). |
schedule-executor | on-demand | on demand | on | Executes ONE scheduled Calendar entry from its dispatch brief — the headless vehicle behind a Command Center Calendar event |
state-validator | recurring | daily 03:00 | on | Daily consistency check — orphan work_claims, routines_state.json drift, KB workspace drift, stale harvest_run rows, schema_version mismatches |
x-content-creator | on-demand | on demand | on | Draft X / Twitter posts (single tweets, threads, replies, polls) from intelligence briefings, news reports, breaking items, and operator-supplied custom files. |
youtube-video-analysis | recurring | 30min | on | Multimodal observer-style routine for the youtube_video pipeline. |