validating-changes

Validates code changes across TypeScript, C++ engine, UI, and audio tiers with smoke tests.

151|22|Updated Apr 19, 2026
One-click install
npx skills add https://github.com/scragnog/HOT-Step-CPP --skill validating-changes-scragnog
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: validating-changes
Source: https://github.com/scragnog/HOT-Step-CPP/tree/main/.claude/skills/validating-changes
Command: npx skills add https://github.com/scragnog/HOT-Step-CPP --skill validating-changes-scragnog

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? It defines the verification bar every change to the HOT-Step CPP three-tier app (C++ engine, Node/TypeScript server, React UI) must clear before being called done, preventing false "done" claims, broken releases, and lost test artifacts. ## Core Features & Use Cases - Per-tier verification bars: TypeScript type-checks, C++ rebuild via dev-rebuild.bat with clean-start log checks, UI visual confirmation by the human, and human-ear audio quality judgment. - Headless end-to-end smoke generation: Submit a generation job through the REST API with auth token, poll status to completion, and confirm WAV output and GENERATION COMPLETED log lines. - Distributability checks: Run check-release-prereqs.mjs to verify model weights, packs, and runtime data actually ship to users, plus a failure-signature table for fast triage. - Use Case: After editing engine source code, rebuild with dev-rebuild.bat, relaunch, confirm the engine log shows Listening on 127.0.0.1:8085, then run a 30-second skipLm smoke generation and hand the WAV path to the human for ear verification. ## Quick Start Ask the assistant to validate the change you just finished by running the appropriate tier checks and a headless smoke generation, then report results honestly.

Frequently Asked Questions about validating-changes

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a headless smoke test for a music generation API?

Check /api/health for engine.ready, verify the queue is free, get a Bearer token from /api/auth/auto, then POST to /api/generate with skipLm true and a short duration. Poll /api/generate/status/:id until it reaches succeeded, failed, or cancelled.

How do I verify a C++ engine change without breaking the running app?

Rebuild only via dev-rebuild.bat at the repo root, which shuts down the app cleanly before compiling. Never call engine build scripts directly, because the Node server auto-respawns ace-server and the respawned process file-locks the linker output.

Can AI judge whether generated audio sounds good?

No. Audio quality can only be judged by a human ear; metrics, file size, spectrograms, and quality-score telemetry have repeatedly failed to predict what listeners hear. Report mechanical completion with the file path, seed, and params, then wait for human feedback.

Why does my generation request return 401 or 503 errors?

A 401 means the Bearer token is missing or stale since tokens are in-memory and reset on server restart, so re-request one from /api/auth/auto. A 503 means the engine is still bootstrapping; wait and check /api/health bootStatus.

How do I check that a feature works for users who download a release?

Run node server/scripts/check-release-prereqs.mjs, which verifies registry model weights exist publicly on Hugging Face, pack file references resolve, and runtime data is packaged by the release workflow. Local presence of files does not prove distributability.