UI はローカライズ済みです。スキルの紹介と説明文書は選択した言語へ AI が自動翻訳し、コード・コマンド・パス・専門用語は原文のままです。原作者の文書が優先されます。
スキルを探す
生産性・共同作業 · SKILL.md

execution-cli

execution-cli

Run, inspect, scale, and troubleshoot Snakemake workflows from the command line, including dry-runs, targets, resources, profiles, batching, grouping, DAG inspection, schedulers, reruns, locks, and smoke checks.

Independent GitHub snapshot · Automated classification · Original documentation governs

スキルの説明

GitHub

name: execution-cli description: "Run, inspect, scale, and troubleshoot Snakemake workflows from the command line, including dry-runs, targets, resources, profiles, batching, grouping, DAG inspection, schedulers, reruns, locks, and smoke checks." disable-model-invocation: true metadata: disco-role: operating license: MIT

Execution CLI

Use this sub-skill when an agent needs to construct or diagnose Snakemake 9.23.1 CLI commands for an existing workflow: dry-runs, target selection, cores/jobs/local cores, resources, profiles, batches, groups, DAG/rulegraph/summary inspection, scheduler tuning, runtime profiling, reruns, locks, and safe smoke checks.

Fast Routing

  1. Confirm the executable and entrypoint with snakemake --help, python -m snakemake --help, or scripts/snakemake_smoke_check.py when the environment is uncertain.
  2. Build a non-mutating command first: explicit --snakefile if needed, explicit targets if requested, --cores 1, --dry-run, and --printshellcmds.
  3. Add scale controls only after the plan is plausible: --cores, --jobs, --local-cores, --resources, --default-resources, --set-threads, --set-resources, and --set-resource-scopes.
  4. Add profiles, batches, groups, rerun/force flags, scheduler flags, or lock cleanup only when the user’s execution goal needs them.
  5. In Snakemake 9.23.1, do not add the legacy --reason flag; dry-run output already includes job-selection reasons.

References

  • Use references/cli-reference.md for option semantics, command templates, profile precedence, DAG/summary commands, scheduler/runtime flags, and API settings names.
  • Use references/execution-patterns.md for end-to-end execution recipes: first run, target runs, profile/resource overrides, batching, grouping, reruns, locks, and smoke checks.
  • Use references/troubleshooting.md for install/import mismatches, missing cores, profile merge surprises, resource-scope errors, stale locks, incomplete outputs, force/rerun misuse, scheduler/ILP issues, and the removed --reason flag.
  • Use scripts/snakemake_smoke_check.py for a safe temporary-workflow probe that does not depend on repository examples or user data.

Boundaries

  • Route Snakefile syntax, rules, checkpoints, wildcards, and shell/script directive authoring to ../workflow-authoring/.
  • Route config files, --config, --configfile, tabular inputs, and sample-sheet semantics to ../configuration-data/.
  • Route conda, containers, executor plugin installation, remote storage, default storage, and deployment packaging to ../deployment-storage/.
  • Route linting, HTML reports, unit tests, notebook/report flags, and deep debugging reports to ../debugging-reporting/.
  • Route Python API equivalents, plugin development, and programmatic embedding to ../python-api-plugins/.

Safety Defaults

Prefer inspectable, reversible commands: dry-run before mutation, explicit targets before broad runs, low cores before scaling, profile overrides only after checking precedence, --unlock only after confirming no Snakemake process is active, and --forceall only when the user explicitly accepts recomputing the reachable workflow.

保存したスキル