Skills catalog

Every agent skill in plugins/robos/skills/, with what it does and how to invoke it. Read this when you want to know which RobOS tasks an agent can already do from a single command.

On this page
  1. How to read this catalog
  2. App development
  3. Knowledge graph
  4. Project plans
  5. Testing, demos and proof of work
  6. Virtual machine and desktop
  7. Agents, review and integrations
  8. Setup and housekeeping

How to read this catalog

Each skill is a directory with a SKILL.md. The descriptions below paraphrase the description line in that file. For the file format, see Skills.

The Invoke column shows one of two things:

  • A slash command, such as /start-vm --headless. These skills have a file in plugins/robos/commands/, and ./plugins/install.sh copies it to .claude/commands/, .antigravity/commands/ and .gemini/commands/. Arguments come from the skill’s Input section.
  • “No command file”. These skills have no slash command. Ask your agent for the task and name the skill, or run the CLI shown in the row.

These are agent skills. The shell commands in the Skills Manager app are a separate list and are on the Skills Manager Desktop App page.

App development

Skill What it does Invoke
create-robos-app Creates and scaffolds a new Electron desktop app, with its registrations and deployment /create-robos-app <app name>
rename-robos-app Renames an app: package directory, desktop entry, icon registry and docs /rename-robos-app <old-name> <new-name>
remove-robos-app Removes an app and deregisters it from manifests, icons and configuration /remove-robos-app <app-id or name>
update-app-icon Replaces or creates the 48x48 Lucide-style SVG icon and syncs it to the icon registries /update-app-icon <app-id or name> [icon description]
add-ai-text-area-to-app Adds the <robos-ai-textarea> widget (auto-resize, streaming, @ file typeahead) to an app /add-ai-text-area-to-app
create-test Generates E2E and unit tests for an app with the robos-test framework /create-test <app-id> <what to test>
create-feature-spec Converts a raw idea or prompt into a feature specification in docs/ideas/specs/ /create-feature-spec <idea text or path under docs/ideas/inbox/>
install-desktop-app Installs and registers a .desktop entry and icon on GNOME Linux, with optional desktop shortcut and dock favorite /install-desktop-app --app-id <id> --name <name> --exec <command>
crpg-game-builder Builds, validates and maintains party-based isometric cRPGs in Godot 4 from the knowledge graph No command file

Knowledge graph

The ten kgraph-* skills wrap the kgraph command-line tool. In this table, kgraph means node packages/robos-graph/bin/kgraph-cli.js. Pass --graph-root <workspace> (or set ROBOS_GRAPH_ROOT) to work on an explicit graph workspace. None of these skills has a command file.

Skill What it does CLI
kgraph-search Searches nodes across all packages by text, RDF type, owner team or tags kgraph search "<query>" [--type <type>] [--package <pkg>] [--team <team>] [--json]
kgraph-insert Adds a node to a package store, gated by shape validation kgraph insert --file <path>, or --id, --type, --title, --package
kgraph-update Changes properties, contracts, dependencies or tags on a node and revalidates it kgraph update <node-id> --set <key=value> or --file <path>
kgraph-delete Removes a node, optionally pruning references to it kgraph delete <node-id> [--cascade]
kgraph-query Queries nodes by type, package or dependency, or traces a path between two nodes kgraph query --type <type>, or --path-from <id> --path-to <id>
kgraph-impact-analysis Traces upstream and downstream dependencies before you change a service, database or contract kgraph impact <node-id> [--depth <n>]
kgraph-validate Validates nodes against the supported built-in shape constraints kgraph validate [package] [--strict] [--json]
kgraph-diff Compares the production main graph with a feature branch or working tree kgraph diff [branch] [--base <branch>]
kgraph-export Exports the graph or one package as JSON-LD, Turtle or N-Triples kgraph export --format <jsonld\|ttl\|nt> [--package <pkg>] [--output <file>]
kgraph-visualize Generates Mermaid or C4 diagrams for a node, package or squad kgraph visualize [id-or-package] [--direction <TD\|LR>] [--output <file>]

kgraph-validate checks supported structure, references and built-in cardinality rules. Its SKILL.md says a passing report is a structural check, not proof that the graph matches the source code.

Other skills that read or write the graph:

Skill What it does Invoke
import-company-kgraph Extracts an evidence-backed graph from local Git checkouts into a reviewable proposal. See Import company knowledge graph /import-company-kgraph with --manifest, --paths, --graph-root, --output
sync-kgraph-docs Inspects graph changes under .robos/, decides what documentation they affect and updates it No command file
schema-lookup Queries Schema.org and OSLC vocabularies, shows type lineage, validates JSON-LD and synthesizes graph shapes No command file. node plugins/robos/skills/schema-lookup/scripts/schema-query.js <search\|type\|property\|validate\|synthesize> ...
generate-app-docs Writes a Markdown architecture page and a Mermaid flow diagram for an app from its graph node, and links them to the documentation package No command file
generate-app-elearning Builds an eLearning course for an app from its graph node, scaffolds a standalone Electron player and records completion certificates No command file

Project plans

These skills read and write saved project plans in a graph workspace. Set ROBOS_GRAPH_ROOT to a directory that contains .robos. The view, update and remove skills also expect ROBOS_SOURCE_ROOT, the absolute path of the RobOS checkout whose CLI you run. The plan format and the headless commands are in packages/task-planner/PROJECT-PLANS.md. The matching UI is the Task Planner app; see its app page.

Skill What it does Invoke
plan-before-implement Creates a reviewed project plan with linked GitHub features and tasks before implementation, and views saved plans by task number or URL /plan-before-implement
view-task-plan Shows or lists saved plans by task number, GitHub issue URL or plan id, without changing data /view-task-plan
update-project-plan Edits, refreshes or restores a saved plan through a reviewed graph proposal, keeping work-item identities /update-project-plan
remove-project-plan Removes one saved plan through a reviewed proposal. The project, tasks, relationships and GitHub issues remain /remove-project-plan

Testing, demos and proof of work

Skill What it does Invoke
e2e-driven-dev Develops a task test-first: writes a text-narrated E2E test, records video with on-screen narration and WebVTT captions, runs on localhost in Docker or Xvfb /e2e-driven-dev <app-id> "<task>", or /do-e2e-driven-dev
record-demo Records a text-narrated walkthrough video of an app on localhost with on-screen narration and WebVTT captions /record-demo <app-id> [what to demonstrate]
test-container Runs RobOS apps in a Docker container with Xvfb and a compositor; runs ./scripts/e2e-container.sh /test-container [--build] [-i] [test file]
app-snapshot Captures a DOM snapshot (text, JSON or screenshot) from a running app, without VNC /app-snapshot <app-id> [--json\|--screenshot]

For how recorded demos are used in review, see Walkthroughs and proof of work.

Virtual machine and desktop

The VM skills drive the QEMU/KVM image built from infra/desktop/. They need QEMU and KVM on the host; install-dev-deps below installs them.

Skill What it does Invoke
build-vm Builds the RobOS QEMU/KVM disk image and the cloud-init provisioning ISO from scratch /build-vm
start-vm Starts the VM; GTK display by default /start-vm [--firstboot] [--vnc] [--spice] [--headless]
stop-vm Shuts the VM down gracefully, or terminates it /stop-vm
vm-status Reports whether the VM runs, SSH reachability, memory, disk, display and installed apps /vm-status
vm-ssh Runs a shell command in the VM over SSH on port 2224 /vm-ssh <command>
deploy-to-vm Copies one or more packages from the repository to the running VM /deploy-to-vm <package...> or /deploy-to-vm all
add-install-step Adds a step to the cloud-init provisioning sequence and the ASCII boot splash /add-install-step <step name>
restart-taskbar Restarts the robos-desktop taskbar and the desktop-manager service in the VM or locally /restart-taskbar

Agents, review and integrations

Skill What it does Invoke
pr-review-theater Inspects and orchestrates PR Review Theater: training knowledge checks, living documentation, in-app diffs and IDE branch comparisons. See PR Review Theater No command file
manage-agent-personas Adds, customizes and inspects developer agent personas and their execution modes /manage-agent-personas <list\|show\|add\|customize\|reset> ...
ide-java Automates Java work in IntelliJ IDEA over the port 63343 bridge: run and debug configurations with secrets from pass, breakpoints, paused-thread inspection /ide-java <subcommand> ...
robos-voice Speaks to the user through RobOS Voice, checks voice status and sends queries to the desktop assistant /robos-voice <speak\|stop\|voices\|assistant\|background\|status> ...
robos-github-auth Rules for code that touches GitHub issues or PRs: use the RobOS task servers and the selected Git client account, never a bare gh call /robos-github-auth

Setup and housekeeping

Skill What it does Invoke
install-dev-deps Audits and installs host dependencies: QEMU/KVM, Electron libraries, Java and Gradle, Node.js packages. Runs scripts/install-dev-deps.sh /install-dev-deps [--check] [--sys] [--npm]
read-error-logs Reads RobOS application failure logs, uncaught exceptions and Electron error-dialog events /read-error-logs [--app <app-id>] [--limit <n>]
report-issue Turns a raw bug report into an issue spec in docs/issues/reported/ /report-issue <bug text or inbox file>
manage-robos-skill Adds, updates, removes or lists skills and syncs them across agent directories /manage-robos-skill <add\|update\|remove\|list> ...