Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add hajibabaie/combinatorial-optimization-skills --skill optuna-hyperparameter-tuninggit clone --depth 1 https://github.com/hajibabaie/combinatorial-optimization-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/hajibabaie/combinatorial-optimization-skills/optuna-hyperparameter-tuning)<a href="https://agentmods.dev/skills/hajibabaie/combinatorial-optimization-skills/optuna-hyperparameter-tuning"><img src="https://agentmods.dev/badge/skills/hajibabaie/combinatorial-optimization-skills/optuna-hyperparameter-tuning.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00137 | $0.12704 |
| Opus 5 | $0.00068 | $0.06352 |
| Sonnet 5 | $0.00027 | $0.02541 |
| Haiku 4.5 | $0.00014 | $0.01270 |
Grade A, and why
optuna-hyperparameter-tuning scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 991 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Optuna Hyperparameter Tuning for Optimization Algorithms
You are an expert in automated algorithm configuration with Optuna. This skill covers tuning the parameters of metaheuristics and MIP solvers: defining search spaces, choosing samplers (TPE, CMA-ES, NSGA-II), building multi-instance objectives, pruning bad configurations early, persisting and parallelizing studies, and validating tuned configurations on held-out instances. Use the pattern catalog below to set up tuning runs that produce parameters which generalize, not parameters that memorize the training instances.
Initial Assessment
Establish these facts before writing any tuning code:
- What is being tuned? A metaheuristic (SA, GA, ALNS), a MIP solver (Gurobi, CP-SAT), or a hybrid? The target determines the objective metric and the cost per evaluation.
- Parameter inventory. List every parameter with its type (continuous, integer, categorical, conditional), a plausible range, and the current default. Fewer than 10 parameters is the normal case; more than 15 suggests the algorithm design should be simplified first.
- Objective metric. Solution quality at a fixed budget? Time to proven optimality? Gap at a time limit? Anytime behavior? The metric must match how the algorithm will be used and reported later.
- Cost per trial. One trial = (instances per trial) x (seeds per instance) x (single-run budget). Compute the total wall-clock cost of the tuning run before starting it.
- Instance set. How many training instances exist? Are they representative of the instances used in the final experiments? Is a train/test split possible (it should be)?
- Stochasticity. Is the tuned algorithm randomized? If yes, plan multiple seeds per instance inside every trial; a single seed makes the objective so noisy that TPE chases luck.
- Tuning budget. How many trials are affordable? Under 30 trials, model-based sampling barely beats random search; plan the budget before choosing the sampler.
- Hardware. Single machine or cluster? Parallel trials need shared storage (SQLite is fine for a handful of workers, JournalStorage or an RDBMS beyond that).
- Reproducibility requirements. Will the tuning protocol be described in a paper? Then fix and record: sampler seed, instance list, seed lists, per-run budgets, and the Optuna version.
- Reporting obligation. Fair comparisons require that every compared algorithm receives the same tuning effort (same budget, same protocol). Decide this now, not after the experiments.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 8d ago First seen · 991 lines · 137 tokens per session scan A fe413fda05ef
optuna-hyperparameter-tuning is a skill published in the GitHub repository hajibabaie/combinatorial-optimization-skills (7 stars, last pushed 2mo ago), licensed MIT. It adds 137 tokens to every session and 12,704 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
semantic-diff
This skill should be used when the user asks to "diff two archetypes", "compare two templates", "what changed between these two versions", "is this a patch, minor or major bump?", "how compatible are these two artefacts?", or invokes /semantic-diff. Auto-detects archetype vs template (any serialisation) and version vs…
bulk-rnaseq
End-to-end bulk RNA-seq orchestrator — takes raw FASTQ reads through QC and trimming (FastQC, fastp/Trim Galore), alignment and quantification (STAR, Salmon, featureCounts), assembles a gene-level counts matrix, then hands off to differential expression (pydeseq2), pathway/GSEA enrichment (pathway-enrichment), and…
genomic-intelligence
Predict regulatory features, gene structure, and expression directly from DNA sequence using Genomic Intelligence's hosted transformer DNA language models — no local GPU or model weights. Six tasks over a REST API and a hosted MCP server (keyless public demo): promoter regions, splice donor/acceptor sites, enhancer…
glycoengineering
Analyze and engineer protein glycosylation. Scan sequences for N-glycosylation sequons (N-X-S/T), predict O-glycosylation hotspots, and access curated glycoengineering tools (NetOGlyc, GlycoShield, GlycoWorkbench). For glycoprotein engineering, therapeutic antibody optimization, and vaccine design.
deepspot-m
Generate transcriptome-wide virtual spatial transcriptomics from H&E histology with DeepSpot-M. Use when you need spatial gene expression in log1p-CPM for 224x224 tiles at about 20x, want to query protein-coding genes by symbol instead of a fixed panel, or want to run prediction across a whole slide after tiling with…
add-ai-webapi
Integrates Power Pages generative-AI summarization APIs (PREVIEW) into a Single Page Application (SPA) site — the Search Summary API and the Data Summarization API — on any record-detail or list page. Generates per-target service code (CSRF-handled) and AI site settings; delegates Web API settings, table permissions…