What problem does it solve?
The skill prevents inconsistent “next-best actions” planning by turning a fuzzy instruction like step back into a standardized, evidence-grounded, time-bounded batch of ranked actions.
Core Features & Use Cases
- Standardized NBA parsing and ranking: extracts count, budget, and focus from natural language and produces a ranked set sized to the time window.
- Mandatory current-state grounding (KB review): checks active PROJECT_PLANs, git status, and session progress so proposals don’t drift or repeat stale work.
- Structural and cross-session safety pre-flights: enforces guardrails for structural skill routing (D71), HANDOFF deferrals (REQ-PA-012), and audit-after-synthesis pairing (REQ-PA-013), while handling aspirational-flag discipline (REQ-PA-014/015).
- Execution-ready output with decision authority: returns a batch ready for the Execute-All default, including per-action Principal classification and batch-level /aget-go gating.
Quick Start
Invoke /aget-propose-actions with your desired count, budget, and focus for this session to receive a ranked set of evidence-grounded next-best actions ready for authorization.