AI Agent Workflow Collection

Use when interrogating the codebase to understand an entity, its fields, relationships, and data flows. Use when user says 'analyze data model', 'analizar modelo', 'understand entity', 'check schema', or asks about a specific entity or database field.

Use when evaluating whether a technical approach is feasible against the current codebase. Use when user says 'analyze feasibility', 'is this feasible', 'evaluar factibilidad', 'technical analysis', or asks whether an approach will work.

Use when extracting user stories from a visual board. Use when user says 'extract stories', 'genera user stories', 'procesa scope', 'board to stories', or shares a board URL.

Use when analyzing a ticket for ambiguities, missing information, or unclear acceptance criteria. Use when user says 'clarify ticket', 'review ticket', 'analyze ticket', or passes a ticket key.

Use when creating a well-structured bug report with root cause analysis. Use when user says 'create bug', 'report bug', 'log bug', 'new bug', or describes a defect.

Use when sending communication to the team. Use when user says 'notify team', 'notifica al equipo', 'send message', 'draft a message', 'enviar mensaje', or needs to communicate with the team.

Establish local connections to cluster services like databases, brokers, and APIs. Use when asked about "port-forward", "connect to database", "local connection", "tunnel", "forward port", "access service locally". TRIGGER when: user asks "port-forward to X", "connect to database locally", "tunnel to service", "forward port", "access service locally", "open local connection to X". SKIP: checking service health (use service-status); running queries (use db-query).

Move failed messages from dead letter queues back to their main queue. Use when asked about "reprocess DLQ", "retry failed messages", "move DLQ messages", "republish dead letters". SKIP: diagnosing failures first (use queue-triage); checking counts (use queue-status).

Check message queue status, list queues, monitor DLQ counts and consumers. Use when asked about "queues", "queue status", "DLQ", "dead letter", "consumers", "messages pending", "message broker", "rabbitmq", "queue health". TRIGGER when: user asks "how are the queues?", "queue status", "any DLQ messages?", "check queues", "messages pending", "consumer count", "queue health", "rabbitmq status", "are there messages in the DLQ?". SKIP: diagnosing DLQ failures (use queue-triage); moving messages back (use queue-reprocess).

Diagnose why messages are failing in dead letter queues and produce a root cause report. Use when asked about "DLQ triage", "dead letter diagnosis", "why are messages failing", "queue errors". SKIP: moving messages back (use queue-reprocess); general counts (use queue-status).

Search and analyze logs from services and containers. Use when asked about "logs", "errors", "exceptions", "log search", "error messages", "stack trace", "application errors", "debug logs". TRIGGER when: user asks "any errors?", "check logs", "what's failing?", "show logs", "stack trace", "exceptions in X", "debug logs", "recent errors", "error messages". SKIP: pod health checks (use service-status); database queries (use db-query); queue failure analysis (use queue-triage).

Check service/container status, health, events, and resource usage across environments. Use when asked about "pods", "service status", "check services", "crashloop", "restart", "container health", "service down", "unhealthy". TRIGGER when: user asks "is X running?", "check pods", "service down", "crashloop", "restart count", "container health", "deployment status", "how many pods?", "pod status". SKIP: log content analysis (use service-logs); database queries (use db-query); queue inspection (use queue-status).

Use when performing a comprehensive ticket analysis across design, documentation, data model, and technical feasibility. Use when user says 'analyze ticket', 'ticket analysis', 'análisis de ticket', or needs deep understanding before implementation.

Interactive deployment workflow with guided intake form and checkpoints at each destructive step. Use when the user wants to deploy a PR or commit with a step-by-step guided experience. TRIGGER when: user asks "deploy PR X", "guided deploy", "walk me through deploying", "deploy workflow", "interactive deploy", "deploy with checks", "step by step deploy", "deploy to prod", "ship PR", "release PR X". SKIP: non-interactive or chained deploys (use deploy directly).