stanfish06 Skills Hub Browser
[!note] Vault audit 2026-07-24 — USE-10 Use this to browse/install bioinformatics community skills from the BioClaw Skills Hub; for discovering skills across the general skill ecosystem use
find-skills. Distinguishing axis: bio/BioClaw domain vs general ecosystem.
Search, browse, and install community-contributed skills from the BioClaw Skills Hub.
The Hub contains 70+ specialized bioinformatics skills organized into domains. Skills downloaded from the Hub are cached locally so they persist for the rest of the session.
When to Use
- User requests an analysis not covered by the built-in skills listed in your system prompt
- User asks "what other skills are available" or "do you have a skill for X"
- User needs a specialized pipeline (e.g., protein design, EHR analysis, spatial transcriptomics workflows beyond the built-in)
Hub Structure
The Hub organizes skills into these domains:
| Domain | Examples |
|---|---|
core-bioinformatics |
alignment-and-mapping, read-qc, sequence-io, database-access |
transcriptomics |
bulk-rna-expression, differential-expression |
single-cell-and-spatial |
scrna-preprocessing, spatial-transcriptomics, cell-annotation |
epigenomics-and-regulation |
atac-seq, chip-seq, dna-methylation |
genomics-and-variation |
variant-calling, genome-assembly, long-read-genomics |
metagenomics-and-microbiome |
metagenomics, phylogenetics, microbial-community |
proteomics-and-metabolomics |
mass-spec, metabolomics |
multi-omics-and-systems |
multi-omics-integration, pathway-analysis |
protein-design |
alphafold2-multimer, proteinmpnn, rfdiffusion, boltzgen (nested under protein-design/skills/) |
ehr-analysis |
electronic health record analysis (single SKILL.md, no per-skill directory) |
How to Execute
Step 1: Fetch the taxonomy (skill index)
curl -sL "https://raw.githubusercontent.com/zongtingwei/Bioclaw_Skills_Hub/main/catalog/taxonomy.yaml"
This returns the full skill catalog organized by domain. Use it to find the skill name that matches the user's need.
Step 2: Resolve the SKILL.md path
Domains do not share one on-disk layout, so never build the path from a template.
protein-design nests its skills under skills/protein-design/skills/<skill>/SKILL.md
and ehr-analysis is a single file at skills/ehr-analysis/SKILL.md, while the other
eight domains use skills/<domain>/<skill>/SKILL.md. Read the real path out of the
repository tree:
curl -sL "https://api.github.com/repos/zongtingwei/Bioclaw_Skills_Hub/git/trees/main?recursive=1" \
| python3 -c "
import json, sys
tree = json.load(sys.stdin)
paths = [t['path'] for t in tree['tree'] if t['path'].endswith('SKILL.md')]
for p in paths:
print(p)
"
Filter that list for the skill you want. To list the skills in one domain, keep the
paths under skills/<domain>/ and take the directory that immediately precedes
SKILL.md.
Step 3: Download and read a skill
# SKILL_PATH is the repo-relative path resolved in Step 2, e.g.
# skills/single-cell-and-spatial/cell-annotation/SKILL.md
# skills/protein-design/skills/alphafold2-multimer/SKILL.md
# skills/ehr-analysis/SKILL.md
SKILL_PATH="<path from step 2>"
SKILL="<skill-name>"
CACHE_DIR="/workspace/group/.hub-skills/${SKILL}"
mkdir -p "${CACHE_DIR}"
curl -fsSL "https://raw.githubusercontent.com/zongtingwei/Bioclaw_Skills_Hub/main/${SKILL_PATH}" \
-o "${CACHE_DIR}/SKILL.md"
curl -f makes a 404 fail loudly instead of caching GitHub's error page as the skill.
Then read the downloaded skill:
read_file({ file_path: "/workspace/group/.hub-skills/<skill-name>/SKILL.md" })
Step 4: Install dependencies (if needed)
Some Hub skills require extra Python packages. Check the SKILL.md for a "Preferred Tools" or "Dependencies" section. Install with:
pip install <package> --quiet 2>/dev/null
Step 5: Execute the skill
Follow the workflow described in the downloaded SKILL.md, just like any built-in skill.
Important Notes
- Always check built-in skills first before fetching from the Hub
- Downloaded skills are cached in
/workspace/group/.hub-skills/for the session - The Hub is a community resource — skills may reference tools not installed in the container; install them with pip/apt as needed
- If GitHub is unreachable, inform the user and suggest using built-in skills instead