
from shinkaevolve1,186
Extracts top-performing programs from a Shinka run and emits a compact Markdown bundle for iteration and analysis.
Shinka Inspect reads Shinka run artifacts (results dir / SQLite DB), ranks programs by combined_score, selects the top-k candidates (with correct-program fallback), and writes a single Markdown context bundle containing metadata, ranking tables, feedback, and code snippets. The artifact is designed to be loaded by coding agents for the next iteration.
Use this skill after an evolutionary run when you need a quick way to surface the best programs for manual review or further automated mutation. It is intended for analysis and preparation before launching further runs or tuning. Avoid when you still need to scaffold or execute runs.
inspect_best_programs.py (see scripts/) to generate the markdown bundle.Best used by coding-capable agents with file-system and python execution access (Copilot-like, local Codex, or CLI-capable agents).
This skill has not been reviewed by our automated audit pipeline yet.