Generate Skill Docs

This documentation was generated with the assistance of AI. Please report any inaccuracies.

Turn the actual, current set of Claude Code skills and agents defined in this repository into this Antora documentation site — the reference table and dependency graph on AI Catalog, the purpose-grouped workflow walkthroughs, and one detail page per skill/agent. This is the skill that generated the page you are reading.

Purpose

Hand-maintaining a page per skill invites drift the moment a SKILL.md changes and its documentation doesn’t. iru-generate-skill-docs closes that gap by discovering the real, current set of skills and agents on disk each time it runs, and regenerating the documentation from them — rather than treating any previously-written page as a source of truth. It only ever writes documentation: it never edits a SKILL.md, an agent definition, or any other source file.

Purpose group

Catalog Tooling

Invocation

/iru-generate-skill-docs

Inputs

Argument Required Description Default

(none)

—

Takes no arguments. Everything it needs — the skill/agent inventory, their dependencies, the purpose groups, the workflows — is discovered fresh from .claude/skills/, .claude/agents/ (if present), and the existing Antora module each time it runs.

Not applicable.

Outputs

  • docs/modules/ROOT/pages/index.adoc updated with a skills-and-agents reference table, purpose-group sections, a Mermaid dependency graph, and one subsection per recurring end-to-end workflow.

  • One detail page per skill under docs/modules/ROOT/pages/skills/<name>.adoc, and per documented agent under docs/modules/ROOT/pages/agents/ (an overview.adoc covering built-in agent types, since this repository has no .claude/agents/ directory of its own to give custom agents individual pages).

  • docs/modules/ROOT/nav.adoc updated with an entry for every new page.

  • No source file, SKILL.md, or agent definition is ever modified.

  • A verified Antora build (npx antora <playbook-file>) confirming every new/updated page and cross-link actually renders before the run is reported as done.

Execution flow

flowchart TD A[Start /iru-generate-skill-docs] --> B{"antora.yml found?"} B -- no --> C["Ask user: bootstrap via\nsetup-antora, or stop?"] C -- stop --> Z1[Stop] C -- bootstrap --> D B -- yes --> D["Discover every SKILL.md,\ncustom agent, and built-in\nagent type actually used"] D --> E["Read + classify each via one\nsub-agent per skill/agent\n(purpose, inputs, outputs,\nstep flow, invokes/spawns)"] E --> F["Build the dependency graph from the\ncollected summaries (invokes / spawns,\nboth directions, external deps noted)"] F --> G["Group skills by purpose"] G --> H["Identify recurring\nend-to-end workflows"] H --> I["Write/update index.adoc\n(table, groups, graph, workflows)"] I --> J["Write/update one detail page\nper skill and per agent"] J --> K["Update nav.adoc\n(add new, remove stale)"] K --> L["npx antora — verify the\nsite builds cleanly"] L -- broken xref/mermaid --> M[Fix and rebuild] M --> L L -- clean --> N[Report]

Dependencies

Invokes

None as a direct Skill() call. Conditionally, if Step 1 finds no Antora module at all, it asks the user whether to bootstrap one via Setup Antora (run through a sub-agent so this skill can keep running afterward) before continuing.

Invoked by

None currently — it’s a top-level entry point, run directly by a person via /iru-generate-skill-docs whenever the catalog’s skill/agent set has changed.

Related agents

  • general-purpose (see Agents) — spawned once per skill and per documented agent in Step 3 (each reading only its own file and returning a compact structured summary, rather than this skill’s own conversation reading every SKILL.md in the catalog directly), and separately in the fallback path to run Setup Antora when no documentation module exists yet.

Source

SKILL.md on GitHub — the file this page was generated from.