Skills
A skill is a reusable bundle of instructions + tools for a specific marketing use case. Skills let the assistant act like a specialist for a narrow task without the org having to manage a separate agent per task.
How skills work#
Each skill has:
- A trigger — when the assistant should activate it, usually a task description or intent pattern.
- Instructions — the prompt override that applies while the skill is active.
- Tool config — which tools are enabled, and per-tool descriptions tuned for this skill’s context.
When the assistant decides a skill applies, it calls select_skill(skill_id) mid-conversation; from that point the active instructions and tool set are the skill’s. The previous turn’s context carries over, so there’s no “reset.”
Built-in skills#
Adlyse ships skills for common tasks: keyword research, creative audit, audience expansion, budget reallocation, competitive analysis, compliance review. Each is scoped to a platform where relevant.
Custom skills#
Create your own from AI Studio → Skills → New Skill. You’ll define:
- Name and one-line trigger
- Markdown instructions (supports the guideline vocabulary —
{{account.name}},{{period.days}}, etc.) - Tool enablement (toggle system tools, custom data-table tools, or MCP-server tools)
- Optional per-tool description overrides
Skills are org-scoped by default. Admins can publish a skill to the org’s marketplace view so teammates see it, or contribute to the Adlyse-wide system catalog (with review).
From MCP#
The skills surface isn’t directly exposed via MCP today — skills are a property of the in-app assistant’s prompt compiler. The underlying tools skills use are available individually through the various MCP servers.