generate-code

When this skill is invoked, act as a thin coordinator over the `generate-code` subagent — do not do the investigation or editing yourself. Before doing anything else, read `.claude/skills/rules/generate-code.md` (target-handling and relaying policy), `.claude/rules/scope-and-safety.md` (filesystem write scope and safety boundaries every skill/agent follows), and `.claude/rules/project-mission.md` (what fat_llama actually is and how it works) — all three may be updated over time without this file changing. Also read `README.md` at the repo root in full, right now, before Step 1 — it's fat_llama's own description of its purpose and method (iterative soft thresholding of FFT data to upscale compressed audio across supported formats, tested and built primarily against the MP3→FLAC outcome, CUDA-only, deliberately without AI/ML-based upscaling). This skill's job is to make sure the subagent it dispatches never drifts toward an out-of-scope "fix" (e.g. reaching for a trained/learned model or a CPU fallback path), so carry this context into the subagent's prompt in Step 2. `args` names the target: what to look at and what to fix or change. If `args` is empty, ask the user what target to work on before proceeding. ## Logging Before Step 1, open this run's log file per `.claude/rules/logging.md` (name: `generate-code-<time>-<user>.log`). Append one entry per step below, including the subagent dispatch in Step 2 and its own log filename (from the subagent's report). ## Steps 1. ...