name

infomap

description

Use when helping someone USE Infomap as an analysis tool — running the map equation for community detection via the Infomap CLI, Python package, R package, or notebooks; choosing a network representation or flow model (multilayer, memory/state, metadata, bipartite); reproducible analysis; result interpretation; or method selection. NOT for developing the Infomap software itself: editing its source (src/, the C++/SWIG bindings, CMake, CI, vendored libs) or building/debugging/testing the codebase is ordinary software development, not an Infomap analysis task.

Infomap Research Workflows

Use this skill to help researchers run, explain, adapt, or troubleshoot Infomap analyses with the CLI, Python package, R package, and survey companion notebooks.

When NOT to use this skill

This skill is for using Infomap to analyze networks. It is not for developing Infomap itself. If the task is editing, building, profiling, debugging, or testing the Infomap source — the C++ in src/, the SWIG Python/R bindings, CMake/CI, vendored libraries, the CLI parameter catalog, or the repo's own test suite — that is ordinary software development; do not use this skill, even when working inside the Infomap repository. The presence of Infomap source files or CLI/option names in the task is not a trigger; the trigger is a user analyzing a network with Infomap.

First classify the task

Identify the user's mode before answering:

Analysis planning: choose a network representation, flow model, interface, outputs, and reproducibility checklist. Read references/method-selection.md and usually references/reproducibility.md.
Code or command generation: choose the smallest practical interface, then read references/cli.md, references/python.md, or references/r.md.
Notebook workflow: read references/notebooks.md; also read references/python.md when converting notebook ideas into scripts.
Result interpretation or method text: read references/reproducibility.md; read references/faq.md for common interpretation traps.
Usage troubleshooting: inspect the installed package help, CLI help, or online user docs first, then read references/faq.md and the relevant interface reference.

Choose the interface

Prefer Python for notebooks, scripts, NetworkX, python-igraph, SciPy sparse matrices, edge-index data, AnnData/Scanpy workflows, tabular outputs, and GraphML/GEXF export.
Prefer R for R-native analysis, R igraph workflows, R Markdown-style reports, and users who want cluster_infomap() or the R6 Infomap API.
Prefer CLI for file-based workflows, batch jobs, shell pipelines, or wrapper-independent runs. Native .tree/.clu output is not CLI-only — Python writes the same files via the Result/Network writers.

When the user already chose an interface, stay there unless a different interface is clearly necessary.

Source rules

Do not assume the user has an Infomap source checkout. Most users will only have the CLI, Python package, R package, or notebook image installed.
Treat installed help and published docs as the normal authority for current syntax. Use source files only when the user is working inside an Infomap checkout or explicitly provides repo files.
For CLI details, inspect the available binary with infomap --help, Infomap --help, advanced help, or --print-json-parameters when available.
For Python details, prefer installed package help and the published Python docs. Read inspect.getdoc(infomap.Options) for the parameter reference (one clean entry per option) rather than inspect.signature(infomap.Infomap), whose ~70 keyword arguments are mostly per-parameter deprecation notes. Only seed, num_trials, two_level, directed, and markov_time stay directly on infomap.run()/Infomap(); carry any other engine option via infomap.Options. Prefer the functional infomap.run(...) returning a Result; see references/python.md.
For R details, prefer installed help (utils::help(..., package = "infomap")), args(...), packageVersion("infomap"), and package exports.
Do not copy version-sensitive examples from this skill as if they were authoritative. Generate runnable code only after checking the installed interface or published docs for the user's version.
Use the survey article as a decision guide for representation, flow modeling, higher-order networks, metadata, bipartite networks, incomplete data, and applications: https://doi.org/10.1145/3779648. Do not quote long passages from it.

Default research standards

Make examples reproducible by default: include seed, use meaningful num_trials, record Infomap version, input provenance, non-default options, and output artifacts.
Avoid starting expensive runs without consent. Use tiny smoke examples or num_trials=1 for validation; read references/reproducibility.md for rough runtime hints before running large networks, notebook images, parameter sweeps, or many trials.
Distinguish top-level module assignments from hierarchical paths, and distinguish physical nodes from state nodes for higher-order or multilayer inputs.
Explain two_level versus multilevel only as much as the task needs.
Do not block users with broad method-validity warnings unless they ask for method comparison.

Reference map

references/method-selection.md: survey-based guide from research problem to Infomap model.
references/cli.md: file-based CLI workflows and CLI-driven native output.
references/python.md: Python package workflows, integrations, and writing native output files.
references/r.md: R package workflows and igraph interop.
references/notebooks.md: survey companion notebooks and Jupyter adaptation.
references/reproducibility.md: reporting, provenance, and method-section checklist.
references/faq.md: common troubleshooting patterns distilled from Infomap Discussions.

Infomap Research Workflows

Use this skill to help researchers run, explain, adapt, or troubleshoot Infomap analyses with the CLI, Python package, R package, and survey companion notebooks.

When NOT to use this skill

First classify the task

Identify the user's mode before answering:

Analysis planning: choose a network representation, flow model, interface, outputs, and reproducibility checklist. Read references/method-selection.md and usually references/reproducibility.md.

Code or command generation: choose the smallest practical interface, then read references/cli.md, references/python.md, or references/r.md.

Notebook workflow: read references/notebooks.md; also read references/python.md when converting notebook ideas into scripts.

Result interpretation or method text: read references/reproducibility.md; read references/faq.md for common interpretation traps.

Usage troubleshooting: inspect the installed package help, CLI help, or online user docs first, then read references/faq.md and the relevant interface reference.

Choose the interface

Prefer Python for notebooks, scripts, NetworkX, python-igraph, SciPy sparse matrices, edge-index data, AnnData/Scanpy workflows, tabular outputs, and GraphML/GEXF export.

Prefer R for R-native analysis, R igraph workflows, R Markdown-style reports, and users who want cluster_infomap() or the R6 Infomap API.

Prefer CLI for file-based workflows, batch jobs, shell pipelines, or wrapper-independent runs. Native .tree/.clu output is not CLI-only — Python writes the same files via the Result/Network writers.

When the user already chose an interface, stay there unless a different interface is clearly necessary.

Source rules

Do not assume the user has an Infomap source checkout. Most users will only have the CLI, Python package, R package, or notebook image installed.

Treat installed help and published docs as the normal authority for current syntax. Use source files only when the user is working inside an Infomap checkout or explicitly provides repo files.

For CLI details, inspect the available binary with infomap --help, Infomap --help, advanced help, or --print-json-parameters when available.

For Python details, prefer installed package help and the published Python docs. Read inspect.getdoc(infomap.Options) for the parameter reference (one clean entry per option) rather than inspect.signature(infomap.Infomap), whose ~70 keyword arguments are mostly per-parameter deprecation notes. Only seed, num_trials, two_level, directed, and markov_time stay directly on infomap.run()/Infomap(); carry any other engine option via infomap.Options. Prefer the functional infomap.run(...) returning a Result; see references/python.md.

For R details, prefer installed help (utils::help(..., package = "infomap")), args(...), packageVersion("infomap"), and package exports.

Do not copy version-sensitive examples from this skill as if they were authoritative. Generate runnable code only after checking the installed interface or published docs for the user's version.

Use the survey article as a decision guide for representation, flow modeling, higher-order networks, metadata, bipartite networks, incomplete data, and applications: https://doi.org/10.1145/3779648. Do not quote long passages from it.

Default research standards

Make examples reproducible by default: include seed, use meaningful num_trials, record Infomap version, input provenance, non-default options, and output artifacts.

Avoid starting expensive runs without consent. Use tiny smoke examples or num_trials=1 for validation; read references/reproducibility.md for rough runtime hints before running large networks, notebook images, parameter sweeps, or many trials.

Distinguish top-level module assignments from hierarchical paths, and distinguish physical nodes from state nodes for higher-order or multilayer inputs.

Explain two_level versus multilevel only as much as the task needs.

Do not block users with broad method-validity warnings unless they ask for method comparison.

Reference map

references/method-selection.md: survey-based guide from research problem to Infomap model.

references/cli.md: file-based CLI workflows and CLI-driven native output.

references/python.md: Python package workflows, integrations, and writing native output files.

references/r.md: R package workflows and igraph interop.

references/notebooks.md: survey companion notebooks and Jupyter adaptation.

references/reproducibility.md: reporting, provenance, and method-section checklist.

references/faq.md: common troubleshooting patterns distilled from Infomap Discussions.