		{
		"name": "@contextium/cli",
		"version": "1.0.31",
		"version": "1.0.32",
		"description": "Command-line tool for managing Contextium documentation, agents, skills, and workflows — pipe context to AI coding assistants",
		@@ -5,0 +5,0 @@ "keywords": [

+61

-31

skills/ium/new-project.md

		@@ -284,47 +284,77 @@ ---

		<step name="regression_agent_question">
		Ask the user:
		<step name="quality_agent_question">
		Before asking this question, classify the project type from the description provided in `project_description`. Choose exactly one from:

		"Would you like a regression testing agent for this project? It will write E2E tests alongside every new feature so nothing you've built ever breaks silently."
		- development — software, apps, APIs, code, engineering, features, bug fixes
		- research — investigation, analysis, literature review, market research, competitive analysis, feasibility studies
		- writing — papers, reports, essays, documentation, articles, books, whitepapers
		- pitch/presentation — pitch decks, slide decks, investor materials, presentations, proposals
		- marketing/content — campaigns, copy, brand materials, social content, creative briefs
		- data/finance — financial models, data analysis, forecasts, reports, spreadsheets
		- design — UX, UI, design systems, wireframes, prototypes
		- other — anything that doesn't clearly fit the above

		Use AskUserQuestion:
		- "Yes, add a regression testing agent"
		Store this as `project_type`.

		Based on `project_type`, determine whether a quality/review agent is relevant and what kind:

		\| Project type \| Relevant? \| Agent name pattern \| What it does \|
		\|---\|---\|---\|---\|
		\| development \| Yes \| "[Project Name] Regression Tester" \| Writes E2E tests alongside every feature so nothing breaks silently \|
		\| research \| Yes \| "[Project Name] Fact Checker" \| Validates claims against sources, flags unsupported assertions, checks citations \|
		\| writing \| Yes \| "[Project Name] Proofreader" \| Reviews drafts for clarity, consistency, grammar, and tone before finalising \|
		\| pitch/presentation \| Yes \| "[Project Name] Deck Reviewer" \| Reviews pitch decks for clarity, narrative flow, and persuasive strength \|
		\| marketing/content \| Yes \| "[Project Name] Copy Reviewer" \| Reviews copy for brand voice, clarity, and effectiveness before publishing \|
		\| data/finance \| Yes \| "[Project Name] Data Validator" \| Checks calculations, flags assumptions, and validates data sources \|
		\| design \| No \| — \| Skip this step entirely \|
		\| other \| No \| — \| Skip this step entirely \|

		If `project_type` is `design` or `other`, skip this step entirely — do not ask the question.

		Otherwise, ask the user the question tailored to their project type. Use natural language that fits the domain — do not use the word "regression" for non-development projects. Examples:

		- development: "Would you like a regression testing agent? It will write E2E tests alongside every new feature so nothing you've built ever breaks silently."
		- research: "Would you like a fact-checking agent? It can review your findings against sources, flag unsupported claims, and validate citations as you build out your research."
		- writing: "Would you like a proofreader agent? It can review drafts for clarity, consistency, and tone before you finalise anything."
		- pitch/presentation: "Would you like a deck reviewer agent? It can review your pitch for narrative flow, clarity, and persuasive strength before you present it."
		- marketing/content: "Would you like a copy reviewer agent? It can check copy for brand voice, clarity, and effectiveness before it goes out."
		- data/finance: "Would you like a data validator agent? It can check your calculations, flag assumptions, and verify your data sources."

		Use AskUserQuestion with:
		- "Yes, add a [agent name]"
		- "No, skip"

		If yes:
		1. Add a "[Project Name] Regression Tester" agent to the creation plan. Its system prompt should:
		- Know that every feature must include an E2E Playwright test as a required deliverable
		- Know the test suite lives in `apps/web/e2e/` and tests run against `https://staging.contextium.io`
		- Know the rule: a feature is not complete until its E2E test is written and the full regression suite passes on staging
		- Be instructed to write focused, realistic tests using existing page selectors and routes — not generic placeholder tests
		2. Store as `regression_agent_planned = true`.
		1. Add the agent to the creation plan. Write a system prompt tailored to the project's specific domain — not a generic description. The prompt should describe:
		- What the agent reviews and what quality bar it holds to
		- What a "complete" deliverable looks like from its perspective (e.g. for development: all features have passing E2E tests; for research: all claims are cited and verified; for writing: no draft goes out without a proofread pass)
		- Any project-specific context inferred from the description (tone, audience, tech stack, subject matter)
		2. Store as `quality_agent_planned = true`, `quality_agent_type = <type>`.

		3. Check if Playwright is already installed by running silently:
		```bash
		cd apps/web && npx playwright --version 2>/dev/null \| grep -q "Version" && echo "installed" \|\| echo "not-installed"
		```
		For development projects only — additionally check if Playwright is installed:
		```bash
		cd apps/web && npx playwright --version 2>/dev/null \| grep -q "Version" && echo "installed" \|\| echo "not-installed"
		```

		If not installed, ask:
		If not installed, ask: "Playwright is required to run E2E tests. Would you like me to install it now?"

		"Playwright is required to run E2E tests. Would you like me to install it now?"
		Use AskUserQuestion:
		- "Yes, install Playwright"
		- "No, I'll install it later"

		Use AskUserQuestion:
		- "Yes, install Playwright"
		- "No, I'll install it later"
		If yes — install silently:
		```bash
		cd apps/web && npm install -D @playwright/test && npx playwright install chromium
		```
		Show `✓ Playwright installed` when done.

		If yes — install silently:
		```bash
		cd apps/web && npm install -D @playwright/test && npx playwright install chromium
		```
		Show `✓ Playwright installed` when done.
		If no — store `playwright_install_pending = true` and add a note to the final summary:
		`⚠ Playwright not installed — run the following before running E2E tests:`
		` cd apps/web && npm install -D @playwright/test && npx playwright install chromium`

		If no — store `playwright_install_pending = true` and add a note to the final summary:
		`⚠ Playwright not installed — run the following before running E2E tests:`
		` cd apps/web && npm install -D @playwright/test && npx playwright install chromium`

		If no: set `regression_agent_planned = false` and continue.
		If no to the quality agent question: set `quality_agent_planned = false` and continue.
		</step>

		<step name="create_resources">
		Only create what the user confirmed across all steps. Create in this order: tag types → tags → libraries → skills → agents (including research, planner, and regression agents if confirmed).
		Only create what the user confirmed across all steps. Create in this order: tag types → tags → libraries → skills → agents (including research, planner, and quality agents if confirmed).

		@@ -331,0 +361,0 @@ Do NOT create a workflow in this step.

@contextium/cli - npm Package Compare versions

New alerts

Fixed alerts

Improved metrics