How to use the scikit-bio agent skill: install and examples

What the scikit-bio skill makes the agent do

An agent skill is a folder of instructions and reference files that an AI agent loads when a task matches it. This skill covers the scikit-bio library, which targets version 0.7.4. Its main file, SKILL.md, describes sequence manipulation, alignment with the pair_align engine, phylogenetic trees, diversity analysis, ordination, and statistical tests such as PERMANOVA.

The skill is documentation and guidance. The agent still writes and runs the Python code itself, using patterns and pitfalls the skill provides.

• Read and write DNA, RNA and protein sequences from FASTA, FASTQ, GenBank and EMBL files
• Run pairwise and multiple sequence alignments
• Build trees with neighbor joining, UPGMA and related methods, and read or write Newick files
• Compute alpha and beta diversity, UniFrac distances, PCoA ordination and PERMANOVA

Install the scikit-bio skill from a ZIP file

Download the ZIP for the scikit-bio skill from this page and unzip it. The skill folder contains SKILL.md and two reference files, references/api_reference.md and references/review.md.

For a personal install, place the unzipped folder in ~/.claude/skills/. To limit it to one project, place it in that project's .claude/skills/ folder instead.

In the Claude apps, you can also upload the skill under Settings → Capabilities → Skills.

• Personal: ~/.claude/skills/scikit-bio/
• Single project: .claude/skills/scikit-bio/ inside your project folder

Set up the Python environment

The skill expects Python 3.10 or newer, scikit-bio 0.7.4 and NumPy 2.0 or newer. The source gives this install command for scikit-bio, which you can run yourself or ask the agent to run:

• Optional: matplotlib and seaborn or plotly for plots
• Optional: biom-format for BIOM tables
• Optional: polars or anndata for table interoperability

Trigger the skill with natural requests

You do not need a special command. Describe the biological task in plain language and mention the file or data you have. The skill is meant to be used when you work with DNA, RNA or protein sequences, file formats such as FASTA or Newick, alignments, trees, or community diversity data.

Name the outcome you want. Asking for 'a Bray-Curtis PCoA plot and a PERMANOVA test on my sample table' is clearer than 'analyse my data'.

• Read sequences.fasta and report the reverse complement of each record
• Align these two DNA sequences globally and show the CIGAR string
• Build a neighbor-joining tree from this distance matrix and save it as Newick
• Compute UniFrac distances for my feature table and run PERMANOVA on the sample groups

Watch for common pitfalls

The skill highlights several choices the agent should make explicitly rather than guess. FASTQ quality offsets should be stated, for example phred_offset=33, rather than inferred from the filename. Translation needs an explicit genetic code and frame, and a decision about incomplete codons.

For alignments, mode='global' allows free terminal gaps by default; use free_ends=False for a fully penalized global alignment. The pair_align_nucl and pair_align_prot wrappers resemble BLAST scoring but are not a BLAST search and report no E-values.

• Ask the agent to state the phred offset and genetic code it used
• Ask it to say whether trees are rooted and which taxa are shared when comparing trees

Safety notes: scripts and network

This skill runs scripts, because the agent executes Python code on your machine. Review the code it proposes before running it on important data. The skill does not use the network, according to its listing.

The examples use small synthetic fixtures. The skill notes that filename-based examples are workflow templates, not validated pipelines, so check outputs against your own reference before drawing conclusions.