Agents, skills and prompts

The Space has one reusable-capability catalog: skills (/space/resources/skills). The prompt library (/space/resources/prompts) and MCP servers (/space/resources/mcp) sit next to it as tabs, and Automations (/space/automations) orchestrates everything that runs on top of them.

New to the AI side? Read AI concepts from scratch first — What makes something an agent, Function/tool calling and Reasoning, which this page assumes.

Agents are skills

There is no separate agent store. Every delegable agent lives in the unified skills registry (src/lib/skills, SKILL.md standard) as a skill with an extra execution profile:

{
  id: 'data-scientist',
  name: 'Data Scientist',
  category: 'data-ml',
  trigger: '/data-scientist',
  instructions: '<agent> … </agent>',   // the SKILL.md payload
  agent: { emoji: '🧮', role: 'Analytics & Machine Learning' },
}

• The instructions field is the agent’s system prompt — the same field a plain skill uses for its procedure. Identity, capabilities, tools and constraints live there in the structured <agent> XML dialect.
• The agent profile is all the “agent” concept adds: an emoji, a role label, optional tools and an optional model override.
• In the UI they appear inside the applied-domain tabs of Resources → Skills (technical agents under Development, comms ones under Writing, business/legal/ops under Finance) — not as a separate section.
• In the chat, list_my_agents returns them alongside the user’s saved agents, and delegate_to_agent accepts both ids. User-saved agents (user_agents table) are the custom-authored complement to the catalog.

The former parallel catalogs (src/lib/agent/library.ts templates, subagents-catalog.ts) were migrated into src/lib/skills/agent-skills-data.ts and deleted. src/lib/agent/library.ts survives only as a compatibility lens derived from the registry.

Automations (formerly the Agent Control Plane)

/space/automations is pure orchestration, n8n-style — three views:

• Pipelines — phased workflow chains from /api/workflows. Run manually from the row.
• Triggers — cron jobs (local manager + server schedules). Create/pause/delete/run-now.
• Runs — unified timeline of pipeline runs + trigger runs, newest first.

The trigger builder targets any delegable entry (catalog skill-agents + your saved agents). The old URLs (/space/studio, /space/hierarchy, /space/cron) permanently redirect to /space/automations. /space/agents had no page of its own and was removed from the router instead of being kept as a redirect.

Skills

/space/resources/skills is the store of reusable capabilities (SkillsPanel, backed by the unified registry). A skill has a name, description, trigger, instructions, category, tags and visibility.

Categories

Office & documents, ecosystem/developer, Hugging Face, Google Labs, Google Workspace, Notion, Stripe, Resend, Figma, Supabase, Cloudflare, GitHub, SEO & growth, and custom skills you author yourself.

Skills complement delegation: instead of editing an agent’s prompt, attach a skill that teaches a repeatable procedure (trigger + instructions) — now literally the same object the agent is made of.

Prompt library

/space/resources/prompts is the collection of reusable prompts (builtInPrompts in prompt-library-panel.tsx). Each prompt has a title, a slash trigger, content, category, tags and visibility.

Built-in triggers

Trigger	Prompt
/review	Code review: performance, security, style, edge cases
/explain	Explain a concept simply + analogy + example
/optimize	Performance optimization with before/after benchmarks
/analyze	Data/metrics analysis with trends, anomalies, takeaways
/debug	Root-cause analysis of a stack trace with fix + diff
/strategy	Executive decision memo (Pyramid Principle)
/saas-metrics	SaaS unit-economics audit (NRR, CAC payback, LTV:CAC, Magic Number, Rule of 40)
/cold-email	3-touch B2B cold email sequence
/legal-review	Contract redlining: liability, indemnification, IP

Prompt categories: strategy, finance & FP&A, marketing & sales, legal & contracts, AI tool prompts, development, analysis, debugging, and My templates for your saved prompts.

How it all fits together

A typical flow: pick a skill from Resources (or let the chat auto-trigger it via its /trigger) → delegate the multi-step part with delegate_to_agent to any entry carrying an agent profile → wire recurring work as an Automation (pipeline or cron trigger) → standardize quality gates with prompts like /review. Everything runs through the same /api/chat streaming pipeline with per-step usage tracking.