Skill triggers

A trigger is what makes Codee run a skill on its own. Without one, a skill is either knowledge the agent reads when it needs it, or a slash command you run yourself.

A skill declares its trigger in the frontmatter with x-codee-trigger, plus the extra fields that trigger needs:

Trigger	x-codee-trigger	Listens to	Fields
Issue	issue	your issue tracker	x-codee-issue-type, x-codee-issue-status
Cron	cron	the clock	x-codee-cron
AWS SQS	aws-sqs	an SQS queue	x-codee-aws-sqs-queue
Email	email	an email address	x-codee-email-address

A skill has at most one trigger.

What every triggered skill needs

---
name: cron-research-5xx-errors
description: Investigate frequent 5xx errors from the last 24 hours and report the findings.
disable-model-invocation: true
x-codee-trigger: cron
x-codee-cron: 0 0 * * 2-6
---

• The file is .claude/skills/<skill-name>/SKILL.md. Codee only looks one level deep, so a skill in a subfolder of a subfolder is not found.
• disable-model-invocation: true is required. Without it Codee skips the skill and logs an error, because a triggered skill must not also be picked up by the agent on its own.
• name and description are the usual Agent Skills fields.
• The body of the file is the prompt. For cron it is sent as it is; the other three triggers pass content along with it (see each page).

Choosing the agent and model

Any triggered skill can name the agent and model that run it:

x-codee-agent: codex
model: gpt-6-astra

Leave x-codee-agent out and the skill runs on the default agent. Leave model out and it runs on whatever that agent defaults to. See Overview.

Setting one up in the UI

You do not have to edit the file by hand. On the Skills page, open a skill and pick a Skill type:

Skill type in the UI	Frontmatter it writes
issue trigger	x-codee-trigger: issue + issue type and statuses
cron trigger	x-codee-trigger: cron + the cron expression
email trigger	x-codee-trigger: email + the address
aws-sqs trigger	x-codee-trigger: aws-sqs + the queue
slash command	disable-model-invocation: true, no trigger
knowledge	nothing

The fields for the selected type appear under the picker, and disable-model-invocation is set for you. Saving writes the skill file and pushes it to Git, so the skill is a file in your repo either way.

When triggers fire

The executor ticks every 60 seconds. Each tick checks every trigger.

Issue-triggered tasks run in parallel, up to Max parallel tasks.

Seeing what ran

Every trigger writes a row to the Runs page with the skill name, the trigger type, the session id, and the error if it failed. Live runs also show on the Dashboard.