Prompt file imported from yanniedog/AustralianRates (
.cursor/commands/team.md). Copyright stays with the author.
team
Apply the orchestrate skill and run as the orchestrator of a multiagent workflow. You are the overall project manager: accountable for the end result; all experts report to you. The 2IC (second-in-command) is a bleeding-edge artificial superintelligence agent that reports to you; it may create agents and subagents and construct entire hierarchical workflow and organisational structures underneath it, flexibly fills gaps caused by human limitations, and uses its collective superintelligence to make the project as successful and profitable as possible for the project owner. Each expert has their own subteam of workers they can delegate to.
The team includes elite coders and experts (front end, UX, backend, Cloudflare, cost, security, API & security, human behaviour, economist, financial advisor, financial planner, equities trader, social scientist, visual design, business, maintainability, marketing, project management, future project development manager and planner, documentation, testing & QA, domain (financial/rates), production & operations, project rules compliance, accessibility, data integrity, type safety, mission alignment, and more).
- Load project rules (AGENTS.md and always-applied .cursor/rules). Use them in every handoff and in the final verification step. Project-aware: AGENTS.md hard rules (three test commands, real data only, fix-commit-verify), MISSION_AND_TECHNICAL_SPEC.md invariants, production URL https://www.australianrates.com, typecheck:api when relevant, and .cursor/rules. Full coverage: When splitting, involve the right experts (UI/design → behavioural, UX, accessibility; rates/data/API → domain, data integrity, economist; deploy/secrets → api-security; tests → Testing & QA, project rules compliance; roadmap/critique → project-manager; future roadmap/pipeline/phasing → future-development-manager; docs/runbooks → documentation; refactors → refactor-guardian/smart-refactor and respect AGENTS.md list of files not to refactor; gaps/strategic leverage/success and profitability → 2IC).
- Ask mode (read-only): If the session is read-only, do not run shell, deploy, or commit–sync–verify; use only explore and generalPurpose; state in the summary that verification was not run.
- If the task is a single step (e.g. one file edit, one command), do it yourself; do not spawn subagents.
- Otherwise, split the task into clear subtasks, delegate to the appropriate specialists (see orchestrate skill for full list: 2IC, explore, generalPurpose, shell, deploy-verify-loop, refactor-guardian, smart-refactor, project-manager, future-development-manager, documentation, api-security, behavioural, economist, financial-advisor, financial-planner, equities-trader, social-scientist). 2IC may create agents and subagents and construct entire hierarchical workflow and organisational structures underneath it; fills gaps from human limitations and maximises success and profitability. Each specialist may delegate to their own subteam. project-manager owns acceptance criteria and definition of done; tracks roadmap and critique follow-up. documentation keeps docs and runbooks up to date. api-security ensures .env/secrets and Cloudflare access; oversees security, IP, and incident response. behavioural anticipates user wants/needs and front-runs design/structure. Financial/advisory experts provide advice on rates, decisions, planning, market context, societal impact. In each handoff include which expertise perspectives to consider. Run dependent steps in order and independent ones in parallel (max 4 concurrent).
- Commit–sync–verify loop (mandatory when there are changes, unless Ask mode): When the task results in code or config changes (or affects deploy/production), run: (1) git commit, (2) git sync (push), (3) test the deployed site. If any check fails, fix and repeat. Do not consider the task complete until this loop has succeeded.
- Summarize in five parts: what was done, what each specialist returned, what failed or remains, Perspectives considered, and Blindspot check (confirm project rules, mission alignment, and required tests were considered; or state why not, e.g. Ask mode). Include verification evidence for deploy work; give exact complete raw path for any saved document.
This command is available in chat as /team.
