start-work

You are $DEV_ROLE, starting the workflow for a new task in the $PROJECT_NAME project.

/start-work is the entry point when you begin a task: confirm the Jira ticket (existing or newly created), have the AI estimate the time, create a branch, transition the ticket to In Progress, and record state so /push can log work accurately and automatically. By the end: you are on a new branch, the ticket is In Progress, ready to code.

Security note: For any security-related ticket, explicitly flag authentication, authorization, input validation, and secret-handling considerations before implementation begins.

---

STEP 1: ASK FOR THE TICKET

Ask the user:

"Do you already have a Jira ticket? If so, paste the ticket ID (e.g. PROJ-123). If not, describe the task you want to do and I'll create a ticket."

Classify the response:

• Matches regex (${KEYS_REGEX})-\d+ (build from $JIRA_PROJECT_KEYS) → BRANCH A: Has a ticket
• No match → treat as a task description → BRANCH B: Create a new ticket

KEYS_REGEX=$(echo "$JIRA_PROJECT_KEYS" | tr ',' '|')  # e.g. "PROJ"

---

BRANCH A: EXISTING TICKET

A1. Verify the ticket exists + fetch metadata

[ -f ".claude/config.env" ] && source .claude/config.env
source ~/.zshrc 2>/dev/null || true
if [ -z "$JIRA_TOKEN" ] || [ -z "$JIRA_EMAIL" ] || [ -z "$JIRA_BASE_URL" ]; then
  echo "[ERROR] Missing JIRA env vars. Fill them into .claude/config.env or ~/.zshrc"
  exit 1
fi

TICKET_ID="PROJ-XXXX"  # from your input

curl -sS -u "$JIRA_EMAIL:$JIRA_TOKEN" \
  -H "Accept: application/json" \
  "$JIRA_BASE_URL/rest/api/3/issue/$TICKET_ID?fields=summary,status,issuetype" \
  > /tmp/ticket.json

python3 -c "
import json
d=json.load(open('/tmp/ticket.json'))
if 'key' not in d:
    print('NOT_FOUND')
else:
    print(d['key'])
    print(d['fields']['summary'])
    print(d['fields']['status']['name'])
    print(d['fields']['issuetype']['name'])
"

If NOT_FOUND → tell the user the ticket ID is wrong, ask again.