version: "1.0.0" name: issue-workflow description: Fetch a GitHub issue and run the full grooming → implementation → review → QA pipeline.

Issue Workflow

Config loading

Read .claude/maestro.json to resolve project-specific values:

• {TEMP_ROOT} = .ai.temp_root
• {REPO} = .ai.repo
• {SLUG} = .ai.slug
• {E2E_URL} = .ai.e2e.local_url
• {E2E_BOOT} = .ai.e2e.boot_cmd
• {E2E_SETTINGS} = .ai.e2e.settings_path
• {E2E_CI} = .ai.e2e.ci_integration

Repository: {REPO} (from .claude/maestro.json)

When the user asks to work on an issue by number, such as:

• /task 123
• issue 123
• #123

follow this workflow. The orchestrator runs inline in this conversation — read the user's opening message before kicking it off, since it uses that for escalation calibration (high autonomy / standard / high oversight).

Tooling

Use shell commands as the primary approach. The GitHub MCP (mcp_github_*) may be used if it is connected — but shell is always the safe fallback and is preferred for reliability.

Steps

1. Extract the issue number from the user's message.

1. Fetch the issue — run bash .claude/skills/issue-workflow/scripts/issue-sync.sh <N> (or use the MCP equivalent). Read the resulting file at {TEMP_ROOT}/issues/<N>/issue.md.

1. Check for parent epics — if Parent Epic (GitHub) or Parent Epics (Task List) has entries, sync each parent with issue-sync.sh <epic-N> and read those files at {TEMP_ROOT}/issues/<epic-N>/issue.md for context.

1. Check if this is an Epic — if the issue has label epics, Issue Type EPIC, or has sub-issues listed, ask the user: "Work the epic as a whole, or a specific sub-issue?" If a sub-issue is chosen, sync it and proceed with the epic context in mind.

1. Determine base branch — default is origin/develop unless the user specified otherwise.

1. Invoke the `orchestrator` skill inline (do not spawn it as a sub-agent — it runs in this conversation context so it can read the user's intent for escalation calibration):

> Inputs: issue number N, issue file {TEMP_ROOT}/issues/<N>/issue.md, base branch

The orchestrator skill manages everything from here: calibration → grooming → spec review → dispatch → implementation → lead review → push & PR → CI → QA → finalize. It spawns the specialist agents (grooming-agent, challenger, backend-agent, frontend-agent, release-agent, lead-reviewer, qa-engineer, ticket-writer) as isolated sub-agents, but the orchestrator itself stays inline so it can surface decisions back to the user naturally.

Monitor progress at {TEMP_ROOT}/issues/<N>/workflow-log.html.