sdlc-triagelisted

# sdlc-triage The non-trivial backlog needs triage: what to pick, what to defer, what blocks what, and which goals can run in parallel? **sdlc-triage** automates the manual flow of surveying the board, asking filter questions, compiling a dependency-sequenced drain plan, and enacting it as native loop primitives — so the plan is durable across sessions, and a later `/sdlc-loop` start picks wave-1's first goal without new integration. The skill is five steps: 1. **Survey** the backlog (fail-soft): inbox (unanswered hand-offs) → active (claimed + in-flight) → parked → enqueued backlog → shadow backlog (assigned, not labelled) → epics (need decomposition **before** picking) → hygiene gaps (missing `priority:*` / `model:*`) → edges (detected dependencies) → context (north-star, wave cap, gates). Detect cross-issue dependency edges (from goal body text and ledger). 2. **Question round** (AskUserQuestion where the host has it; plain questions otherwise): pick / defer / drop per triage bucket, then per-issue refinements. Confirm detected edges + ask for missing ones. Confirm the wave capacity (default from `parallel.goals.max_concurrent`). Fill missing `priority:*` and `model:*` labels on picked issues (collected here, applied by enact). The `Bash(gh issue *)` grant covers reading goal issue bodies during this phase to detect edges. 3. **Compile** the plan via `triage.py plan`: topo sort over union(detected edges, user edges); cycle → hard error namin