SH

shinka-inspect

Extract and package top-ranked Shinka programs into a markdown summary for agent review.

Install

mkdir -p .claude/skills/shinka-inspect && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/16007" && unzip -o skill.zip -d .claude/skills/shinka-inspect && rm skill.zip

Installs to .claude/skills/shinka-inspect

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Load top-performing Shinka programs into agent context using `shinka.utils.load_programs_to_df`, and emit a compact Markdown bundle for iteration planning.
155 charsno explicit “when” trigger
Intermediate

Key capabilities

  • →Load top-performing Shinka programs into a DataFrame.
  • →Rank programs by `combined_score`.
  • →Select a specified number of top correct programs.
  • →Generate a Markdown bundle with metadata, ranking table, and code snippets.
  • →Truncate program code snippets to a maximum character limit.
  • →Filter programs by a minimum generation number.

How it works

The skill uses `shinka.utils.load_programs_to_df` to read program records, ranks them by `combined_score`, and generates a Markdown bundle with selected program details.

Inputs & outputs

You give it
A path to a Shinka run results directory, e.g., `--results-dir <results_dir>`
You get back
A Markdown bundle file, e.g., `<results_dir>/shinka_inspect_context.md`, containing top programs.

When to use shinka-inspect

  • →Inspecting top programs after an evolution run
  • →Creating context files for next-step planning
  • →Comparing code snippets from successful runs

About this skill

Shinka Inspect Skill

Extract the strongest programs from a Shinka run and package them into a context file that coding agents can load directly.

When to Use

Use this skill when:

  • A run already produced a results directory and SQLite database
  • You want to inspect top-performing programs before launching the next batch
  • You want a compact context artifact instead of manually browsing the DB

Do not use this skill when:

  • You still need to scaffold a task (shinka-setup)
  • You need to run evolution batches (shinka-run)

What it does

  • Uses shinka.utils.load_programs_to_df to read program records
  • Ranks programs by combined_score
  • Selects top-k correct programs (fallback to top-k overall if no correct rows)
  • Writes one Markdown bundle with metadata, ranking table, feedback, and code snippets

Workflow

  1. Confirm run artifacts exist
ls -la <results_dir>
  1. Generate context bundle
python skills/shinka-inspect/scripts/inspect_best_programs.py \
  --results-dir <results_dir> \
  --k 5
  1. Optional tuning knobs
python skills/shinka-inspect/scripts/inspect_best_programs.py \
  --results-dir <results_dir> \
  --k 8 \
  --max-code-chars 5000 \
  --min-generation 10 \
  --out <results_dir>/inspect/top_programs.md
  1. Load output into agent context
  • Default output path: <results_dir>/shinka_inspect_context.md
  • Use it as the context artifact for next-step mutation planning

CLI Arguments

  • --results-dir: Path to run directory (or direct DB file path). In a run directory the script reads programs.sqlite, or the database the run wrote under a custom db_path or a first-release name (evolution_db.sqlite). If the directory holds several program databases under other names, it stops and asks for the file path.
  • --k: Number of programs to include (default 5)
  • --out: Output markdown path (default under results dir)
  • --max-code-chars: Per-program code truncation cap (default 4000)
  • --min-generation: Optional lower bound on generation
  • --include-feedback / --no-include-feedback: Include text_feedback blocks

Notes

  • Ranking metric is combined_score.
  • If no correct rows exist, script falls back to top-score rows and labels fallback in output.
  • Script is read-only for run artifacts (writes only the markdown bundle).
  • With a shinka installed from PyPI at 0.0.8 or earlier, which has no shared database lookup (a source checkout can report 0.0.8 too and have it: check that shinka.database.run_db_path imports), the script reads <results_dir>/programs.sqlite, or evolution_db.sqlite when only that exists, without checking that the file holds programs. With neither, it takes the only other .sqlite or .db file in the directory (not prompts.sqlite) and stops when there are several: pass the database file path then.

When not to use it

  • →When the user needs to scaffold a task using `shinka-setup`.
  • →When the user needs to run evolution batches using `shinka-run`.
  • →When a run has not yet produced a results directory and SQLite database.

Limitations

  • →The skill requires a run to have already produced a results directory and SQLite database.
  • →The ranking metric is `combined_score`.
  • →The script is read-only for run artifacts, only writing the markdown bundle.

How it compares

This workflow automates the extraction and bundling of top-performing Shinka programs into a structured Markdown file, providing a compact context artifact for coding agents, unlike manual browsing of the database.

Compared to similar skills

shinka-inspect side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
shinka-inspect (this skill)08moReviewIntermediate
biorxiv-database71yReviewBeginner
office-docs19moReviewBeginner
catalog-extractor04moReviewIntermediate

Try saying

Example prompts that trigger this skill in your AI assistant.

Search skills

Search the agent skills registry