Recipe: Slack notifier hook¶

Problem¶

You want the team to know when someone's environment is up. Maybe to welcome a new contributor, maybe to flag drift, maybe just so the lead engineer can sip coffee and watch onboardings tick by. Send a Slack message at the end of jarvy setup — without making it spam or blocking the install on Slack outages.

Config¶

jarvy.toml

[hooks]
post_setup = """
# Skip in CI — only notify on real laptops.
if [ -n "${CI:-}" ]; then exit 0; fi

# Slack webhook URL is per-team; sourced from env, never hardcoded.
if [ -z "${JARVY_SLACK_WEBHOOK:-}" ]; then exit 0; fi

USER_NAME="${JARVY_USER:-${USER:-someone}}"
HOST_NAME="$(hostname -s)"
PROJECT="$(basename "$PWD")"

curl -s -X POST -H 'Content-Type: application/json' \\
    --data "{\\"text\\":\\":wave: $USER_NAME finished \\`jarvy setup\\` for \\`$PROJECT\\` on $HOST_NAME\\"}" \\
    "$JARVY_SLACK_WEBHOOK" \\
    || true   # never fail setup on a Slack outage

exit 0
"""

[hooks.config]
continue_on_error = true   # advisory: setup succeeds even if Slack 500s
timeout = 10

Why it works¶

Piece	What it does
`if [ -n "$CI" ]; then exit 0; fi`	CI runs don't notify — only laptop runs do.
`if [ -z "$JARVY_SLACK_WEBHOOK" ]; then exit 0; fi`	If the webhook isn't set, the hook is a silent no-op. Developers who haven't opted in get nothing.
`\\|\\| true`	Slack down? Setup still succeeds. The notifier is advisory.
`continue_on_error = true`	Belt and suspenders — even if curl errors, setup keeps going.
`timeout = 10`	A misbehaving Slack endpoint can't hang setup forever.

Each developer sets JARVY_SLACK_WEBHOOK in their shell rc:

~/.zshrc

export JARVY_SLACK_WEBHOOK="https://hooks.slack.com/services/T0/B0/xxxx"

Or stash it in 1Password and pull it via op — see Secrets via 1Password.

Variations¶

Notify only on first setup, not re-runs:

[ -f .jarvy/state.json ] && exit 0   # already set up, skip

Notify on drift instead of setup:

[hooks]
post_setup = """
DRIFT=$(jarvy drift check --format json 2>/dev/null || true)
if echo "$DRIFT" | grep -q '"drifted": true'; then
    curl ... "drift detected on $USER's $HOST"
fi
"""

Discord / Teams / Mattermost: same pattern, different webhook URL and JSON shape. Most chat platforms accept { "text": "..." }.

Channel routing by project:

case "$(basename $PWD)" in
    frontend-*)  CHANNEL="#frontend-onboarding" ;;
    backend-*)   CHANNEL="#backend-onboarding" ;;
    *)           CHANNEL="#general-onboarding" ;;
esac

Then post to a Slack app that routes by channel name in the JSON.

Track time-to-setup as a metric:

DURATION=$(jq '.duration_seconds' .jarvy/state.json)
curl ... "$USER finished setup in ${DURATION}s"

Caveats¶

Webhook URLs are secrets. Don't commit them. Use the env-var pattern.
Don't notify in CI. New runners on every PR = Slack channel firehose. Always guard with if [ -n "$CI" ]; then exit 0; fi.
Rate limits matter. Slack incoming webhooks cap at 1 message/sec. If your hook fires for many tools, batch into one message.
Privacy: the message includes $USER and hostname. If your team is privacy-conscious, drop those.