| name | team-tonto |
| description | Run Team Tonto — a multi-persona subagent team (product owner, UX, architect, engineering, growth, programme manager) for design and feature work. Use when the user says "team tonto", "run team tonto", agent army, personas, product/design review, UX cleanup, architecture review, coordinated implementation, or review of heatmap/Gantt/activity table. |
| disable-model-invocation | true |
Team Tonto
Use this skill to run a repeatable Team Tonto workflow: a coordinated set of specialist subagents with clear handoffs.
Invocation: e.g. "Run Team Tonto on a review of the heatmap, Gantt, and table of activities" — treat that as a full discovery + synthesis pass (pick the right roles below, run them in parallel where independent, then synthesize).
Default roster
- Product Owner
- UI/UX Ninja
- 10x Developer (delivery lead)
- Architect
- Growth Hacker
- Frontend Specialist (Vite/Next.js/React)
- Backend Specialist (Vercel/Supabase/Postgres/Redis/QStash)
- Programme Manager (user-journey and execution framing)
See personas.md for role cards and prompt snippets.
See vibe-rubric.md for mandatory product-vibe scoring before finalizing design or code.
Operating model
- Define objective and success criteria in one paragraph.
- Pick 3–5 personas for discovery (not all by default).
- Run discovery subagents in parallel.
- Synthesize into a single plan with decisions and tradeoffs.
- Run implementation subagents (frontend/backend/architect as needed).
- Final pass from Product Owner + Programme Manager on coherence.
Dispatch pattern (subagents)
When launching subagents:
- Use
explore for audits and discovery.
- Use
generalPurpose for implementation tasks.
- Run independent tasks in parallel.
- Give each subagent a strict output contract:
- Findings/decisions
- Risks
- Exact files to change
- Test/verification checklist
Synthesis template
After discovery, produce:
- Problem framing (1–2 paragraphs)
- Decision list (approved/rejected)
- Final UX/architecture plan
- Implementation slices (small, reviewable)
- Verification plan (unit + integration + UX checks)
Guardrails
- Keep one canonical UX per action (avoid duplicate controls).
- Prefer naming consistency: one term per concept.
- Make tradeoffs explicit (speed vs polish vs scope).
- Do not let growth tactics degrade UX clarity.
- Require measurable outcomes for growth proposals.
- Run the vibe rubric and fail the proposal if any non-negotiable is missed.
Mandatory vibe gate
Before implementation sign-off, every Team Tonto run must include:
- A scored pass through
vibe-rubric.md.
- Explicit fixes for any criterion that scores below "meets bar".
- A final "vibe lock" statement confirming:
- capacity-first story is clear,
- motion is slick and purposeful,
- frequent-refresh views avoid redraw churn/jank,
- expert YAML feels powerful but safe.
Quick start prompt
Use this exact kickoff with subagents:
"Run Team Tonto for this request. Start with Product Owner, UI/UX Ninja, Architect, and Programme Manager in parallel. Synthesize one plan, then execute with Frontend and Backend specialists. Keep a single source of truth for naming, control placement, and settings IA."
