Produce two prompts the user can paste into agent sessions. Does not implement code.
Invoke (no slash required)
The user may say:
Generá issue-slice-prompts para #67Prompts zoom-out + tdd del issue #62- Attach this skill manually: @issue-slice-prompts or Skills picker → issue-slice-prompts
There is no guaranteed /issue-slice-prompts command unless you register it in Cursor Settings → Rules, Skills.
Process
1. Load the slice
gh issue view <N> --repo SantiagoXOR/pintureria-digital --json number,title,body,labels,url
If body has ## Parent, fetch parent for PRD context. Respect Blocked by.
2. Light codebase exploration
Grep/read real modules, routes, tests. Use tenant / embudo / Resumen vocabulary.
3. Prompt A — zoom-out
/zoom-out
Contexto: Issue #<N> — <title> (parent #<P>). <1 línea>
Tarea zoom-out:
1. Mapea capas y callers (<entrada> → UI/API/DB>) con rutas reales del repo
2. Bugs/deuda del PRD si aplica
3. Módulos a crear vs modificar
4. Diagrama + tabla módulo → responsabilidad → callers
No implementes. Interfaces públicas + 3–5 comportamientos testeables prioritarios.
4. Prompt B — TDD
/tdd
Implementa issue #<N> con TDD vertical (un test → cambio mínimo → GREEN).
Antes de codear, confirma interfaz pública y orden de tracer bullets.
Reglas (desde AC): <bullets>
Tracer bullets:
1. RED→GREEN: <behavior> — <test + prior art>
...
Out of scope: <...>
Prior art: <src/__tests__/...>
One behavior per bullet. Split TRACK A/B when issue allows parallel work. No horizontal "all tests then all code".
5. Footer
- Attach:
zoom-out,tdd, domain skill if needed - Branch:
feat/issue-<N>-<slug>
Output
- Links (issue, parent, blockers)
- Prompt A (fenced)
- Prompt B (fenced)
- Session order if multiple issues
Chain
to-prd → to-issues → issue-slice-prompts → zoom-out → tdd
See examples.md.