Agent: Skills, Subagents, Workflows, Rules, Memory

The Agent tab is where you customize AI behavior. These files aren't for you to read — they're for the agent to read. They define how it should work inside your project.

This tab is hidden by default

Enable it under Settings → Agent. Leaving it off costs you nothing day to day — Scholardo ships a sensible set of defaults.

The five sections

Section	What it is	Stored in
Skills	Reusable task patterns / prompts. The agent invokes them when a task matches	agent/skills/
Subagents	Reusable role definitions. Workflows can reference them	agent/subagents/
Workflows	The definition of a multi-subagent orchestration: how to decompose, delegate, and synthesize	agent/workflows/
Rules	Behavioral norms for the agent	agent/rules/
Memory	Knowledge entries	Project / system level

Memory is hidden by default as well, and enabled in the same place.

Project vs System scope

Each section's contents split across two scopes:

• Project — applies to this project only, travels with it
• System — applies to every project, stored under your user directory

Right-click any entry to Move to Project / Move to System.

The test is simple: does this rule still hold in a different project? "Always cite in APA" is System-level. "Sample IDs in this project follow XX-YYYY" is Project-level.

Skills get a structured editor

Markdown files under agent/skills/ open in a dedicated editor that also maintains the skills index, so renaming a skill or changing its description keeps the agent's roster in sync.

Files in the other sections open in the regular Markdown or code editor.

Workflows are definitions, not an executor

An easy point to misread: Scholardo authors and manages workflow definitions, but execution belongs to the CLI agent. A workflow file describes how the main agent should decompose a task, which subagents to delegate to, which models to use, and how to merge results — the thing that actually runs is the agent in the right pane.

Agents can read their own assets

Agents can inspect their own capabilities over MCP:

• scholardo_list_agent_assets — list available skills / subagents / workflows
• get_rules — read the rules currently in effect
• list_memory / read_memory / save_memory / update_memory — read and write memory entries

So you can simply ask "what skills do you have right now," or tell it to "save what we learned here to memory."

Session overview

The top of the Agent tab shows how many agents are currently running ("N active"). With none detected it reads "No agents detected" — start a session in the right pane.

Going deeper

• Agent & MCP Tools — the full tool list, connection, and live resources