The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill directs users to install Nextflow and the nf-test framework by fetching scripts from https://get.nextflow.io and https://get.nf-test.com and piping them directly to bash. This execution method is flagged as a risk when originating from non-whitelisted domains.
[COMMAND_EXECUTION]: The skill facilitates the creation and execution of shell scripts within Nextflow processes that interpolate variables from external sources, potentially allowing for arbitrary command execution if the input is malicious.
[PRIVILEGE_ESCALATION]: Setup instructions include the use of sudo to move the Nextflow binary into /usr/local/bin/, which involves granting administrative permissions for file system operations.
[EXTERNAL_DOWNLOADS]: The skill downloads pipeline code, configurations, and container images from various third-party sources, including GitHub (nf-core/*) and container registries like quay.io and galaxyproject.org.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through its handling of external data.
Ingestion points: The skill ingests untrusted data from CSV samplesheets via the splitCsv operator in references/language.md and references/running-pipelines.md.
Boundary markers: No explicit delimiters or instructions are used to separate ingested data from shell script templates.
Capability inventory: The skill allows for the generation of shell scripts, file writing, and container execution.
Sanitization: The instructions lack validation or sanitization steps for input data before it is interpolated into shell commands.

nextflow