Scaffold a new strategy (and optionally environment) model in the plant C++/R package, then wire it through the build. Use when asked to add a new model/strategy, create a strategy from an existing one (e.g. clone FF16), add a variant strategy that reuses an existing environment, or "use the scaffolder". Covers running scripts/new_strategy_scaffolder.R, the own-environment vs reuse-environment modes (issue #274), implementing the biology, and the make rebuild + test loop. Does NOT cover changing the numerics of an existing model.
Migrate code that uses the plant R package up to plant's latest interface. Use when a downstream product (analysis scripts, a dependent package, a notebook) breaks or needs updating after plant changed — e.g. "update X to the latest plant", "plant's interface changed, fix my code", "migrate to the new run_scm/Control API", "bring this repo up to date with plant". Reads plant's NEWS.md "Breaking changes" as the spec, finds affected call sites, and proposes per-site fixes. Does NOT change plant itself or invent API mappings — if NEWS lacks a needed mapping, it stops and asks.
Profile and benchmark the plant C++/R package — find hotspots, measure whether a change or branch helped or regressed, and compare performance across builds. Use when asked to profile, benchmark, time, or optimize plant; to find what is slow; to compare speed between two branches/commits; or to verify a change did not regress timing. Covers the build-then-load workflow, the sample-off same-machine A/B methodology (machine drift is large — only same-session ratios are trustworthy), native /usr/bin/sample hotspot analysis (Rprof alone is too coarse for C++), and the timing-history reporting convention.