Guide · Agent skill
How to use the pacsomatic Tumor-Normal Workflow Helper agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The pacsomatic skill is an add-on for an AI agent that prepares and launches the nf-core/pacsomatic pipeline for a matched tumor and normal pair of PacBio HiFi genome sequencing files. This guide explains what the skill does, how to install it, how to ask for it in plain language, and what you need before you run it. By the end you will know how to get a reviewed dry run of your pipeline files and when to launch the job.
pacsomatic Tumor-Normal Workflow HelperK-Dense-AI/scientific-agent-skills · MIT Download the skill Quick start
- 1Download and unzip. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder.
- 2Check requirements. Confirm Python 3.10+, Bash, Nextflow 24.04.2 or newer, a compatible Java runtime and a container runtime are available.
- 3Describe the run. Ask the agent for a pacsomatic run, naming the tumor and normal BAMs, patient ID, sample IDs, output directory and reference.
- 4Generate a dry run. Have the agent run the helper with --dry-run to write the samplesheet, params YAML and script without launching anything.
- 5Review the artifacts. Check the samplesheet, params file, script, pipeline revision, profile and any warnings before proceeding.
- 6Launch or submit. When the review is complete, ask for --run, using --executor for a scheduler if needed.
- 7Check the result. Confirm the submission ID or run output and plan the next QC step, since scheduler acceptance is not pipeline completion.
What the pacsomatic skill makes the agent do
Agent skills are folders of instructions and scripts that an agent loads when a task matches them. This one covers a narrow job: a matched tumor and normal PacBio HiFi run using the nf-core/pacsomatic Nextflow pipeline, starting from unaligned BAM files.
The agent uses the bundled helper script to check inputs, generate a samplesheet and reproducible launch files (a params YAML and a shell script), and then either stop at a dry run or launch the Nextflow driver locally or through a scheduler. It also helps troubleshoot startup problems.
- Checks tumor and normal BAM paths, patient ID, sample IDs and exactly one reference option (--fasta or --genome)
- Writes a samplesheet with one normal and one tumor record
- Writes a params YAML and a launch script for review
- Launches locally or submits to LSF, Slurm, PBS Pro or SGE
- Helps diagnose startup failures
Install the pacsomatic skill
Download the skill ZIP from this page and unzip it into one of these locations. Use ~/.claude/skills/ for a personal install available across projects, or a project's .claude/skills/ folder to limit it to one project.
If you use the Claude apps instead, upload the ZIP under Settings → Capabilities → Skills.
- Personal: unzip into ~/.claude/skills/
- Project: unzip into .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP
Ask for it with natural requests
You do not need a special command. Describe the job in plain language and name pacsomatic or the tumor/normal HiFi workflow. The skill is meant for pacsomatic run preparation and execution, so requests about short-read somatic calling or hospital imaging systems will not match it.
Good requests name the tumor and normal BAM files, the patient ID, the reference, and whether you want a dry run or a launch.
- "Prepare a pacsomatic dry run for tumor P001_T.bam and normal P001_N.bam with the GRCh38 FASTA."
- "Show me the samplesheet and params file before submitting anything."
- "Why did my pacsomatic Nextflow launch fail at startup?"
Example: dry run first, then a Slurm launch
The skill's usual flow is to generate artifacts with --dry-run and review them before running. The example below is from the skill's documentation; paths are illustrative placeholders. Run it from the repository root.
After you review the artifacts, a Slurm launch can reuse the same inputs and swap --dry-run for --run. The driver resources are separate from task resources, and the Slurm config must already set the task executor.
- Dry run does not invoke the pipeline, validate BAM contents, or check remote availability
- Slurm walltime 48:00 is normalized to 48:00:00
- Do not add a sanger profile unless you use that institution's LSF setup
python skills/pacsomatic/scripts/run_pacsomatic.py --tumor-bam /data/P001_T.bam --normal-bam /data/P001_N.bam --patient-id P001 --tumor-sample-id P001_T --normal-sample-id P001_N --outdir /results/P001 --fasta /refs/GRCh38.fa --profile apptainer --use-current-path --dry-runWhat you need before running
The skill checks files but cannot install your software. Load cluster modules before invoking the helper, because --module-load only repeats commands in the generated script. Network access is needed for uncached pipeline code, plugins, references and containers.
- Python 3.10 or newer (standard-library helper)
- Bash
- Nextflow 24.04.2 or newer
- A compatible Java runtime (Java 17-26)
- A container runtime such as Apptainer or Docker, as selected by the profile
- Optional scheduler command: bsub, sbatch or qsub
- Local HiFi BAM inputs (optional .pbi indexes) and a reference FASTA or genome key
Safety notes: scripts and network
This skill runs scripts and uses the network. Read the generated script and params file before any --run or --submit. Existing artifacts require --overwrite, and input files are never overwritten.
Use managed filesystem or cloud credentials for remote URIs. Do not put secrets or signed URLs in generated files. Confirm PacBio HiFi read groups, sample identity and reference compatibility yourself; the skill does not verify biological suitability. The upstream work is a source-reviewed development workflow, not a clinically validated assay.
- Do not use short-read data with this skill
- Scheduler acceptance is not pipeline completion
- Keep the output directory stable so --resume works
Tips for smoother runs
Start with --dry-run every time and read the warnings. Keep patient and sample IDs free of whitespace and make sure the tumor and normal files are different. Confirm that any .pbi index belongs to its BAM, and that MM/ML modification tags survived processing if you need methylation results.
Keep paths inside external params and config files absolute, because the launcher runs from the output directory.
- Use --use-current-path or an existing --conda-env to select the runtime
- config.yaml is an operator reference and is not loaded automatically
- Read references/config-and-output.md for scheduler and output details
What people use it for
- Preparing a matched tumor and normal PacBio HiFi run on an HPC cluster with reviewable launch files
- Reproducing a pacsomatic launch from saved params and script files for a research group
- Reviewing a dry run before an operator submits a pipeline job to Slurm, LSF, PBS Pro or SGE
- Diagnosing a Nextflow startup failure for a pacsomatic run
Frequently asked questions
Is the pacsomatic skill free to use?
The skill is published under the MIT licence, as stated on this page. The pipeline and your compute, storage and container resources are separate and depend on your own setup.
What do I need to install before using pacsomatic?
You need Python 3.10 or newer, Bash, Nextflow 24.04.2 or newer, a compatible Java runtime (Java 17-26) and a container runtime such as Apptainer or Docker. A scheduler command (bsub, sbatch or qsub) is needed only for cluster submission.
Where do I put the pacsomatic skill folder?
Unzip the ZIP into ~/.claude/skills/ for a personal install, or into .claude/skills/ in a project. In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.
Can pacsomatic analyse short-read sequencing data?
No. The skill is scoped to pacsomatic and unaligned PacBio HiFi BAM files. Short-read or general somatic analysis is outside its scope, and using the wrong platform gives wrong results.
Does the skill run clinical diagnostics?
No. The skill states that it is a source-reviewed development workflow and not a clinically validated assay. It is not for clinical diagnosis or patient care, and it is unrelated to hospital PACS imaging archives.
Why did my pacsomatic run fail at startup?
Ask the agent to troubleshoot the startup. Common causes covered by the skill include missing runtime tools, Java or Nextflow versions, container profile problems and unreachable remote inputs, which should be checked against the requirements list.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use the pacsomatic skill if you are a bioinformatician or research group preparing matched tumor and normal PacBio HiFi runs with nf-core/pacsomatic, especially on an HPC cluster, and you want reviewable launch files and a dry run first. Pick something else for short-read somatic analysis, general cancer genomics, or any clinical or hospital imaging work.
Download the skill