Skip to main content
This is the Desktop workflow. The IDE extension is a separate product; see the IDE Extension docs.
A Task is the container for a piece of work. It holds agents, files, the git diff, and artifacts: specs, tickets, stories, and reviews. Traycer ships a set of skills that tell agents how to plan, build, and review through those artifacts.

​
How Artifacts Appear

Artifacts are written when something should outlive the conversation or be put in front of you. You can create them from the Artifacts panel, and an agent can write one during a turn. An agent writes an artifact only when the work calls for one. If none is written, the coding work can still finish and the context stays in the chat transcript. To be sure you get an artifact, ask for it, or run the skill that produces it.

​
Built-In Skills

For a large feature, a typical order is brief, core flows, tech plan, tickets, execute, review, and walkthrough. Skip any step the work doesn’t need.

​
Plan

​
Build

​
Review

​
Decide and Maintain

​
Where Skill Output Lands

Folders are relative to the Task’s artifacts. See Artifacts on Disk.

​
Run a Skill

In the chat composer, type $ or / to open the picker; both open the same list. Pick a skill, then write your prompt after it. You can attach several skills to one prompt.
Some coding agents also start a skill on their own when a request matches its description. Traycer doesn’t run the steps as an automatic pipeline and doesn’t fail a turn when a skill is skipped. Picking the skill yourself is the reliable way to run it. To add skills of your own, see Skills.

​
Feedback Through Comments

Comment on the artifacts as you review them, then ask an agent to address the open threads. The agent reads each thread, updates the artifact, and resolves the threads it has handled. See Comments and Agents.