Command reference

Every command is a command_<name> function in src/skillgen/cli.py, registered in HANDLERS. This page lists every command and flag. For each command's --json output see JSON output.

Shared options

Every command accepts:

Option	Effect
--json	Print JSON, validated against src/skillgen/schema/cli/<command>.schema.json before it is printed
--root FOLDER	Repository root (default: the current folder, or the nearest parent holding .skillgen/)
-v, -vv	More output on the console: info, then debug. The log file always gets everything

skillgen --version prints the version.

Exit codes

Code	Name	Meaning
0	EXIT_OK	The command ran and everything passed
1	EXIT_FAILED	The command ran but something failed (a check, a stale skill, a missing tool)
2	EXIT_ERROR	An expected error stopped it (a SkillgenError: bad config, unknown route, export refused…)
3	EXIT_NEEDS	The run stopped at a next step for the agent or developer; see next

Set up

init

Create .skillgen/ with its config.json, routes.json, rules/, pages/ and a .gitignore for secrets/, cache/ and logs/.

Flag	Effect
--app-name "My App"	The app's display name; also the default skill-name prefix
--url URL	Base URL of a running dev or staging copy, for --xy
--environment dev|staging|production	What --url points at. The probe refuses production
--name PREFIX	Prefix for every skill name
--analyzer claude-code|file|replay	Who performs the analysis step

skillgen init --app-name "My App" --url http://localhost:3000

install

Add /skillgen, the route-analyzer subagent and the driver skill to Claude Code, and download Chromium for --xy when the probe extra is installed.

Flag	Effect
--target FOLDER	Claude folder (default ~/.claude; use ./.claude for one project)
--no-chromium	Skip the Chromium download

doctor

Check Python, Graphify, Claude Code, ast-grep and Playwright, and where skillgen runs from. Exits 1 when a required check fails.

Flag	Effect
--fix	Download Chromium for the probe

login

Open a browser window so you can log in to your dev copy once. The session is saved in .skillgen/secrets/storage-state.json for probing. Needs the probe extra.

Flag	Effect
--url URL	Login page (default: the app URL)

Find pages

routes

Find routes from every source and write .skillgen/routes.json. See Route discovery.

Flag	Effect
--all	Every route (the default)
--route /x	Add a route by hand; repeatable
--include GLOB	Route glob to keep, such as 'campaigns/*'; repeatable
--exclude GLOB	Route glob to drop; repeatable
--name PREFIX	Prefix for every skill name
--no-graph	Do not build the code graph
--ask-claude	Also let Claude Code list routes from the code graph's likely router files
--no-claude	Never call Claude Code, even when no route is found

Make skills

generate

Generate one skill per route. Exits 3 when a route waits for a next step. See Generating skills.

Flag	Effect
--route /x	One route, a concrete page, or its skill name. Exclusive with --all
--all	Every route in routes.json
--name NAME	With --route: the exact skill name. With --all: the prefix for every skill
--xy	Also measure where each field sits on the dev copy; only adds
--url URL	Base URL of a running dev or staging copy; implies --xy
--open STEP	With --route: what to click before the form shows, a CSS selector or text:Label; repeat for several; saved to routes.json
--code-only	Never open a browser (the default)
--analyzer claude-code|file|replay	Who performs the analysis
--replay-dir FOLDER	Recorded analyses for --analyzer replay
--concurrency N	Routes analysed at once
--description "…"	Use this skill description instead of the generated one
--include GLOB, --exclude GLOB	Keep or drop routes; repeatable
--force	Also retry routes marked as having no form
--overwrite	Recreate skills that already exist; without it they are left as they are and reported
--no-graph	Skip the code graph; slice entry files only

skillgen generate --route /campaigns/new --name new-campaign
skillgen generate --all --name myapp --xy
skillgen generate --route /ads/manager --name ad-budget --xy --open "text:Ad sets" --open "text:Edit budget"

prepare

Write the slice and the analysis prompt for one route, so Claude Code can analyse it in your session. /skillgen runs it for you.

Flag	Effect
--route /x	The route (required)
--name NAME	The skill name for this route
--no-graph	Skip the code graph

Answer and refine

questions

List facts skillgen needs you to confirm.

Flag	Effect
--skill NAME	Only this skill's questions
--all	Include answered questions

answer <id> <answer>

Record an answer: yes, no, code, page, or a value. It applies on the next generate. See Questions and rules.

skillgen answer q1a2b3c "text:Audience settings"

add-rule [skill] "<rule>"

Add a rule to one skill's user block, or to every skill with --global.

Flag	Effect
--check "KEY CONDITION"	A machine check compiled into fill.js, such as 'utm_campaign ^[a-z0-9_]+$' or 'daily_budget min 10'
--global	Add to .skillgen/rules/global.md for every skill

skillgen add-rule new-campaign "Daily budget must be at least 10" --check "daily_budget min 10"
skillgen add-rule "Never launch a campaign without a UTM tag" --global

Check and share

validate [skill…]

Check skill folders against the Agent Skills spec and skillgen's rules, and scan them for secrets. With no names, every skill is checked. Size notes are reported but never fail.

index

Write skills-index.html (graph and skill cards) and skills-index.json at the repository root.

check

List skills made stale by code changes. Exits 1 when something is stale. See Keeping skills current.

Flag	Effect
--since REF	Git reference to compare with (default HEAD~1)
--regenerate	Regenerate the stale skills
--replay-dir FOLDER	Recorded analyses, when the analyzer is replay
--no-graph	Match changed files against slices only

export

Copy skills out for sharing, after checks and a secret scan, with MANIFEST.sha256. See Exporting.

Flag	Effect
--zip	Write one zip file
--one-skill	Bundle every page into one skill with one SKILL.md, for an upload; written to dist/skill/<name>
--plugin	Lay out as a Claude Code plugin
--name NAME	Plugin or skill name (default: the app's prefix)
--out FOLDER	Destination (default dist/<name>)

export-chatgpt

Pack every page into one ChatGPT skill (one SKILL.md plus a file per page, with plugin.json and an icon), after the same checks and scan.

Flag	Effect
--zip	Write one zip file, ready to upload
--name NAME	Plugin and skill name (default: the skill prefix)
--out FOLDER	Destination (default dist/chatgpt/<name>)

When something goes wrong

logs

Show the latest run log, or a list. Every command writes one to .skillgen/logs/.

Flag	Effect
--list	List recent logs instead
--lines N	Lines of the latest log to show (default 60)
--count N	Logs to list (default 10)

probe

Probe one route on a running dev copy and save what it saw. Only for looking; it does not change a skill. Needs the probe extra.

Flag	Effect
--route /x	The route (required)
--url URL	Concrete page URL, for routes with parameters