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 agentmods add skills/yale-som-hpc/claude-code-marketplace/parallel-pythonnpx skills add yale-som-hpc/claude-code-marketplace --skill parallel-pythongit clone --depth 1 https://github.com/yale-som-hpc/claude-code-marketplaceWhat 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 | $0.00061 | $0.03852 |
| Opus 5 | $0.00030 | $0.01926 |
| Sonnet 5 | $0.00012 | $0.00770 |
| Haiku 4.5 | $0.00006 | $0.00385 |
Grade A, and why
parallel-python 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 2d 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 — 350 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Parallel Python
Rule: pick one main layer of parallelism, match it to the bottleneck, and size workers from the Slurm allocation. If one process is too slow, first check accelerating Python.
Choose the right tool
| Workload | Use | Avoid |
|---|---|---|
| Independent tasks across files/years/specs | Slurm array or GNU parallel | One giant Python process |
| CPU-bound pure Python | ProcessPoolExecutor or multiprocessing |
Threads, because of the GIL |
| I/O-bound HTTP/API/database work | threads or async | Too many processes |
| Variable-duration tasks | imap_unordered or as_completed |
Static equal chunks |
| Large numeric kernels | BLAS/NumPy/Polars/DuckDB threads | Extra Python processes unless needed |
| GPU training/inference | one process per GPU pattern | CPU preprocessing inside GPU allocation |
Respect the Slurm allocation
SLURM_CPUS_PER_TASK is set inside Slurm jobs and unset on your laptop. Use a small helper so the same script works in both places:
import os
def available_workers() -> int:
"""Match the Slurm allocation when running under sbatch/srun, fall back
to the local machine's CPU count for laptop / login-node testing."""
slurm = os.environ.get("SLURM_CPUS_PER_TASK")
if slurm:
return int(slurm)
# cgroup-aware fallback. Do NOT use os.cpu_count(): inside a small allocation
# it returns the whole node (e.g. 32) — verified on this cluster — so a job
# with SLURM_CPUS_PER_TASK unset would oversubscribe massively.
return len(os.sched_getaffinity(0))
If Slurm gives you 4 CPUs, do not start 32 workers. If each worker calls BLAS/NumPy, also set thread environment variables in the Slurm script.
Avoid nested parallelism
This is the single biggest waste pattern on the cluster.
Bad: 500 Slurm array tasks × 16 Python processes × 8 BLAS threads = 64,000 runnable threads.
Good: 50 concurrent Slurm tasks × 1 Python process × 1 BLAS thread; or: 1 Slurm job × 16 Python processes × 1 BLAS thread
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.
- 2d ago First seen · 350 lines · 61 tokens per session scan A cb035aed64bd
parallel-python is a skill published in the GitHub repository yale-som-hpc/claude-code-marketplace (5 stars, last pushed 1mo ago), licensed Unlicense. It adds 61 tokens to every session and 3,852 once invoked, about $0.0003 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
bump-dependency
Bumps a Python package dependency across Home Assistant Core integrations, regenerates core requirement files, runs verification tests and prek lint, and prepares a pull request with proper release/compare links.
biopython
Comprehensive molecular biology toolkit. Use for sequence manipulation, file parsing (FASTA/GenBank/PDB), phylogenetics, and programmatic NCBI/PubMed access (Bio.Entrez). Best for batch processing, custom bioinformatics pipelines, BLAST automation. For quick lookups use gget; for multi-service integration use…
marimo-pair
Work inside the user's live marimo notebook from the code editor: run Python in the same kernel the user does, inspect live notebook state, and commit durable notebook changes through code mode. Use whenever you create, analyze, or improve the user's marimo notebook.
adk-style
Python style and codebase conventions for ADK (Agent Development Kit): private-by-default file visibility, imports, type hints, Pydantic v2 models, formatting, docstrings, logging, async I/O, file and test layout, and unit test structure. Use when writing or editing ADK source or tests, deciding whether a new file or…
coding
编写并运行 Python 代码，验证脚本逻辑和输出。.
ax-python-agent
Use when writing Python code with axllm for agents, child delegation, tools, MCP, citations, persistent playbook learning, stage instructions, runtime state, final typed responses, and direct-respond executor skipping.