version: "1.0.3" name: rai-bugfix-close-lean description: "Lean residue de rai-bugfix-close: push + MR + Jira sync. 10 pasos vs 26 enterprise. Asume pipeline-orchestrated." model: haiku

allowed-tools:

• Read
• "Bash(git:*)"
• "Bash(glab:*)"
• "Bash(rai:*)"

license: MIT metadata: raise.adaptable: 'true' raise.fase: '5' raise.frequency: per-bug raise.gate: '' raise.skillset: raise-maintainability raise.version: 1.0.0 raise.visibility: public raise.work_cycle: bugfix-lean raise.inputs: |

• bug_id: string, required, argument
• run_id: string, required, injected by orchestrator in pointer prompt
• dev_branch: string, required, config

raise.outputs: |

• mr_url: string, terminal

Bugfix Close (Lean)

⚠ LEAN VARIANT (ADR-134 v2) — si cambias lógica [R]/[G] en rai-bugfix-close, propaga aquí.

Fuente: work/epics/e14770-lean-methodology/skill-audit.md §rai-bugfix-close

Lean residue: 10 pasos [R9, G1, I2] de 26 en enterprise (54% ceremonia eliminada).

Asume siempre pipeline-orchestrated — requiere run_id en el pointer prompt.

Pipeline Context Check

Verificar que este skill fue invocado vía pipeline engine (RAISE-10715):

• Si Run ID: aparece en el contexto → continuar silenciosamente.
• Si ausente → STOP: "standalone execution detected — prior phases may have been skipped. Options: 1) continue anyway, 2) abort."
• Si eliges continue anyway → ANTES de continuar, emite señal de observabilidad del bypass (RAISE-15391 — el "continue anyway" antes no dejaba rastro queryable):

``bash rai signal emit-work bug "<bug_id>" --event blocked --phase close \ --blocker "standalone-close: prior phases (review/retro) unverified" ``

Objetivo

Push del branch de bug, crear MR, y sincronizar Jira. El lean pipeline siempre provee run_id — no hay check de contexto de pipeline. Sin ceremony.

Inputs: Bug ID, run_id (inyectado por el orquestador), {dev_branch}.

Pasos

1. Verificar artifacts via rai pipeline status [R]

rai pipeline status "{run_id}"

Construir mapa {phase_id → status}. Para cada artifact:

• scope.md (fase: start/diagnose) → si fase skipped: ok; si fase ran + archivo ausente: STOP
• analysis.md (fase: analyse) → si fase skipped: ok
• plan.md (fase: plan) → si fase skipped: ok
• retro.md (fase: review) → si fase skipped: ok

En lean, analyse/plan/review siempre son skipped → solo scope.md es obligatorio.

2. Verificar árbol limpio [R]

git status --short

Si hay cambios uncommitted → commitarlos antes de push.

3. AR gate [R]

rai gate check gate-ar-bugfix

Escape hatch para XS/docs:

RAISE_AR_SKIP_REASON="<razón>" rai gate check gate-ar-bugfix

ℹ El gate emite automáticamente una señal ar-skip (WorkLifecycle, queryable vía rai signal query bugfix <branch> --event ar-skip) al usar el escape hatch — RAISE-15391. Es un rastro de observabilidad además del governance/ar-skip-log.jsonl. No la re-emitas manualmente (doble conteo).

Si falla sin escape hatch → STOP: el AR debe existir antes del close.

4. Alembic single-head gate [R] (si el fix incluye migraciones)

rai gate check gate-alembic-single-head

Si falla → la cadena alembic tiene fork. Linealizar down_revision antes de push — no hacer merge sin esto; alembic upgrade head fallará en deploy.

Si el bug no toca migraciones → este gate pasa automáticamente (skipped).

5-6. Push + Crear MR [G]

Write the title and human-readable description to files without shell interpolation. Invoke /rai-mr-create with these required inputs:

• source_branch: current bug/{issue_key}/{slug} branch
• target_branch: {dev_branch}
• work_id: {bug_key}
• quality_review_file: work/bugs/{bug_key}/qr.md
• title_file: file containing the exact MR/PR title
• description_file: file containing the exact human-readable body

GitLab uses THIN admission. GitHub retains the local full-gate path until equivalent remote enforcement exists. Neither provider permits manual push or MR/PR creation outside /rai-mr-create.

If /rai-mr-create fails (sync conflict, gate failure, push rejection), STOP. Do not assign fixVersion, advance the pipeline, or clean up. There is no remote branch/MR result to finalize: returning a successful close would allow the pipeline to transition Jira despite an unvalidated commit.

7. fixVersion — delegada al engine [R]

No asignar fixVersion manualmente — el engine lo hace automáticamente en el siguiente pipeline_advance: tras la transición a done y antes de after:bug:close, finalize_post_transition escribe el mismo fixVersion que verifica gate-close-jira-sync (assign_close_fix_version, resuelto desde la release line destino del branch — RAISE-18975). Asignar desde el branch name produce un conflicto con gate-close-jira-sync (RAISE-16614).

8. Verificación post-transición delegada (blocking) [R]

No ejecutar gate-close-jira-sync directamente desde el skill: Jira todavía no está en Done. Al devolver control, el siguiente pipeline_advance aplica primero target_status: done y después ejecuta el gate en after:bug:close.

Si el avance devuelve gate_failed para after:bug:close → FAIL el close. No log-and-continue ni escape implícito. Corregir Jira y re-run.

9. Worktree-aware cleanup [R]

if [ "$(git rev-parse --git-common-dir)" != "$(git rev-parse --git-dir)" ]; then
  echo "Worktree — cleanup diferido a /rai-worktree-close"
else
  git checkout {dev_branch}
  git branch -D bug/{issue_key}/{slug}
fi

10. Bind result [I]

rai session bind RAISE_SESSION_JIRA_KEY "{bug_key}" 2>/dev/null || true

Fin. Devolver: MR URL, gate-close-jira-sync status, cleanup realizado sí/no.

Backlog tracker

Engine behavior (RAISE-16987): The engine checks ExecutionContext to determine mode:

• mode: pipeline — engine active; apply_phase_transition moves the tracker. No skill call required.
• mode: standalone — no active run; tracker is not moved automatically.

If standalone (mode == standalone via ExecutionContext):

rai backlog transition {key} --point after:bugfix:close