Agent Activation

Agents are the working personas you activate for a task. This is different from skills:

  • agents are active workers or specialists
  • skills are instruction packs you load when needed

This guide is about the user-facing agent workflow: inspect agents, activate the right ones, understand dependencies, and know when to let recommendations help.

Core Commands 

cortex agent list
cortex agent status
cortex agent activate code-reviewer
cortex agent deactivate code-reviewer

Useful inspection commands:

cortex agent deps code-reviewer
cortex agent graph
cortex agent validate

Three Ways Agents Get Activated

1. Manual activation

Use this when you already know which specialist you want:

cortex agent activate security-auditor
cortex agent activate python-pro code-reviewer

This is the most direct path and the easiest one to reason about.

2. AI recommendations

Use the agent intelligence system when you want Cortex to suggest who should be active based on the current repo context:

cortex ai recommend
cortex ai auto-activate

This is recommendation-driven activation, not skill loading.

3. Dependency-driven activation

Some agents pull in related dependencies. Use the dependency tools to understand what an activation implies:

cortex agent deps architect-review
cortex agent graph

Activation Workflow

The practical flow is:

  1. check what is active now
  2. inspect the current task
  3. activate the smallest useful set of agents
  4. deactivate agents that are no longer relevant

Example:

# Inspect current state
cortex agent status

# Activate the specialists you need
cortex agent activate python-pro code-reviewer

# Later, trim back the set
cortex agent deactivate python-pro

Deactivation

Deactivation matters because an overgrown active set can add noise.

cortex agent deactivate code-reviewer
cortex agent deactivate architect-review --force

--force exists because dependency checks can block deactivation when other agents still rely on the one you are removing.

TUI Workflow 

cortex tui

Then:

  • press 2 for the Agents view
  • inspect active versus available agents
  • activate or deactivate directly from the interface
  • use 0 to compare against AI recommendations in the AI Assistant view

When To Use Agents vs Skills

Use agents when:

  • you want a specialist persona active for the task
  • you want recommendation-driven activation from cortex ai
  • you are managing the working set of collaborators

Use skills when:

  • you want reusable instructions or references
  • you want prompt-time or recommender-driven skill suggestions
  • you want a context file for another session or model