xtarterize

name

xtarterize

description

Apply production-grade conformance configurations (linting, type checking, CI, editor settings, AI skills, release tooling) to JS/TS projects. Scans package.json and config files to auto-detect framework, bundler, and styling, then resolves applicable tasks. Use when user mentions "xtarterize", "conformance", "setup linting", "add CI", "configure project", "init xtarterize", or any task name like "lint/biome", "ts/strict", or "ci/workflow".

xtarterize

Scans any JS/TS project and applies curated configs via a task-based engine. Always use --format json for agent-driven invocations.

When this skill loads

This skill is activated when the user asks about project conformance, linting setup, CI, or any xtarterize task. Upon loading:

1. Assess the ask - is the user requesting a full setup (init), a single task (add), a status check (check), a task search (query), or diagnostics (doctor)?
2. Check current state first - run check --format json to understand what's already configured before proposing changes
3. Never guess task IDs - load the task reference before using add
4. Always use --format json - parse structured output, never terminal text

Quick reference

npx xtarterize check --format json   # { tasks: [{ id, label, group, status }] }
npx xtarterize diff --format json    # FileDiff[]
npx xtarterize list --format json    # { profile, tasks }
npx xtarterize query "strict typescript" --format json  # { type: "query", results: [...] }
npx xtarterize doctor --format json  # { diagnostics: [{ name, status, message }] }
npx xtarterize add lint/biome --format json
npx xtarterize init --format json --yes

JSON parsing patterns

After each command, parse the JSON to decide the next action:

Command	Parse this field	To decide
check --format json	tasks[].status	If any status is "new" or "patch", run init or add
check --format json	diagnostics[].status	If any is "fail", run doctor before proceeding
diff --format json	Array length	If [], nothing to change. If non-empty, show diffs to user
diff --format json	[].filepath	Which files will be modified
diff --format json	[].stats.added + removed	Magnitude of changes per file
list --format json	tasks[].status	Find the right task ID to pass to add
list --format json	profile	Understand the detected stack (framework, bundler, etc.)
doctor --format json	diagnostics[].status	Any "fail" needs fixing; "warn" is advisory
init --format json	ok	If false, something went wrong - check stderr
query --format json	results[].relevance	Score >= threshold means relevant. Use results[].signals to see which fields matched strongest
query --format json	results[].taskId	Pass to add to apply the matched task
query --format json	count	If 0, no tasks met the threshold - broaden query or lower --threshold

All commands

Command	Purpose	Flags (always add --format json)
init	Apply all applicable tasks	--yes, --dry-run, --skip <ids>, --only <ids>, --include-conflicts, --compose <query>, --cwd <path>
sync	Update existing configs	Same flags as init
diff	Preview pending changes	--cwd <path>
check	Audit conformance per task	--verbose, --cwd <path>
add <id>	Apply one task	--cwd <path>
list	List tasks with statuses	--cwd <path>
query <query>	Search tasks by natural language	--limit <n>, --threshold <n>, --cwd <path>
restore <file>	Recover from backup	Interactive only
doctor	Environment diagnostics	--verbose, --cwd <path>

Task statuses: "new" (will create), "patch" (will update), "skip" (conformant), "conflict" (needs review - require --include-conflicts).

Agent workflows

Initialize a project

# 1. Check state
out=$(npx xtarterize check --format json --cwd "$dir")
# Parse out.tasks[].status - if all "skip", skip init

# 2. Preview
out=$(npx xtarterize diff --format json --cwd "$dir")
# Parse out - if empty array, nothing to do

# 3. Apply
out=$(npx xtarterize init --format json --yes --cwd "$dir")

Add one task

# MANDATORY - Load references/tasks.md first to find exact task ID
out=$(npx xtarterize add ts/strict --format json --cwd "$dir")

Diagnose failures

out=$(npx xtarterize doctor --format json --cwd "$dir")
# Parse out.diagnostics - any { "status": "fail" } needs attention
# Common fails: "Biome not installed", "Node version too old"

Error handling

Signal	Likely cause	Fix
add <id> → "Task not found"	Wrong task ID	list --format json, parse tasks[].id
doctor shows tool not installed	Missing dep	xtarterize only writes configs; user may need to install separately
Task shows "conflict"	Config differs	Never auto-apply - present to user, only with --include-conflicts
--cwd fails preflight	No package.json	Verify path exists and is a JS/TS project
query returns empty results	Query too narrow or no matching tasks	Broaden query, try synonyms, or lower --threshold

Anti-patterns

• NEVER run init --yes without diff --format json first on existing projects
• NEVER edit .xtarterize/backups/ - use restore to recover
• NEVER use --include-conflicts without user approval
• NEVER parse terminal output - always use --format json
• NEVER guess task IDs - load references/tasks.md first

Reference files

• Load references/tasks.md before add - find exact task IDs. MANDATORY - READ ENTIRE FILE.
• Load references/commands.md for detailed flag descriptions and JSON output shapes. Do NOT load for basic usage - covered above.