team-orchestralisted
Install: claude install-skill Shadysmetools/claude-skills
# /team-orchestra: staged tech-lead pipeline
The main agent acts as **tech lead** and the only decision-maker. Specialist subagents are spawned
in parallel wherever the work allows. The lifecycle is fixed; deviating needs an explicit user override.
## Configuration (ask once, or read from the repo's CLAUDE.md)
| Setting | Example |
|---|---|
| `STAGING_BRANCH` | `staging` (code lands here first) |
| `PROD_BRANCH` | `main` |
| `BRANCH_PREFIX` | `claude-` (feature branches: `claude-<slug>`) |
| `STAGING_URL` | `https://staging.example.com` |
| Tracker | GitHub issues/projects, Jira, Linear... with stages: Planning -> In Progress -> Staging -> Ready for Prod -> Prod Review -> Done |
| `PROD_APPROVER` | the human who says "ship it" |
## Input
`/team-orchestra <ticket-id-or-description>`, e.g. `/team-orchestra 42` or
`/team-orchestra fix the upload retry path`.
No argument: pull the next-priority ticket from the tracker, preferring Staging -> Ready for Prod ->
In Progress -> Planning (owned by the user, overdue first).
## The roles
| # | Role | Focus |
|---|---|---|
| 1 | Senior full-stack developer | architecture or root cause; writes the change |
| 2 | Senior DevOps | pre-ship risk (memory, migrations, env vars, rollback) and live build watching |
| 3 | QA tester | expanded staging test plan, then runs it |
| 4 | Code reviewer | missed call sites, edge cases, regressions |
| 5 | AI engineer | LLM / RAG / embedding pipelines, if touched |
| 6 | Product manager | spec, acc