Package and catalog

Privacy policy

The harness-first plugin does not collect data for stark AI. This notice covers the package, this catalog, and what stays with the host or GitHub.

Contact supportReport a security issue

Who we are

servrox solutions UG publishes this catalog and the stark AI Developer package under the stark AI brand. For privacy questions, email dev@stark-ai.de.

The responsible person is: Aaron Röhl. Provider: servrox solutions UG (haftungsbeschränkt), Panoramastrasse 26, 71296 Heimsheim, Germany. See the Imprint for full identification. The stark AI product website privacy notice covers that marketing site, not this catalog.

What this notice covers

It covers the harness-first plugin and this static catalog. It does not describe processing by ChatGPT, Codex, Claude Code, Cursor, GitHub, GitHub Pages, an installer, or any other host, workspace, repository, or tool. Those services use their own notices.

The package

The package is Markdown skill instructions, OpenAI skill metadata, references, scripts, templates, and static assets. It has no publisher-operated backend, MCP server, account system, telemetry, or analytics.

When a host uses a skill, that client or model may see the prompt, files, workspace metadata, command output, or other artifacts you make available. The package does not send prompts, repositories, files, outputs, or credentials to stark AI.

Jev Capability Advisor has an optional TypeSafe API connection. When you request a fresh recommendation, its local helper sends your task text and public capability cards to TypeSafe and authenticates with your key. It does not execute recommended tools or send their results. Offline candidate inspection makes no API request. Only supply task text and catalog descriptions you intend to share with that provider; its processing is governed by its own terms and privacy notice.

Optional local caches and JSON receipts may contain task text, capability metadata, and recommendation results. They are stored in the paths you choose and remain on your machine until you remove them. The helper excludes authentication headers from its receipts. Do not put credentials in task text or catalog descriptions.

Telegram Serverless guides use of Telegram's official CLI and hosting platform. When you authorize deployment or remote execution, the CLI sends selected modules, built frontend files or test inputs to Telegram. Remote execution can access the bot's database, send Bot API messages and call external services used by your application. Linking uses the owner's separate Serverless CLI access token, entered locally. Installation of this skill alone does not contact Telegram or create a bot. Telegram and any application-selected services process that content under their own terms and privacy notices; it is not sent to stark AI by this package.

Answers for this package

These answers apply to the harness-first plugin. They do not describe a host product, GitHub, or this site's hosting provider.

  • Publisher collection: no package telemetry is sent to stark AI. Task text sent to TypeSafe and bot data or test inputs sent to Telegram may contain personal data.
  • Purpose of service processing: TypeSafe recommends capabilities; Telegram links, runs and hosts the bot or Mini App you authorize. Optional local advisor caches reuse recommendations; receipts support inspection.
  • Recipients: TypeSafe receives enabled advisor requests; Telegram receives authorized CLI operations and deployed application traffic. Your application may call other services you select. This package does not route that content to stark AI.
  • Retention: you control local output and cache files. TypeSafe, Telegram, your application services and your host set their own service retention policies.
  • Your controls: use Support for ordinary questions and Security for suspected vulnerabilities. Email dev@stark-ai.de for privacy requests about this package or site.
  • Host and product processing: ChatGPT, Codex, the selected host, workspace, repository, installer, and tools may process content under their own policies. This package cannot guarantee those processors.
  • Accounts, backend, and telemetry: the package has no stark AI account, backend, telemetry, analytics, or hidden network calls.
  • Support and security reports: if you email us or open a GitHub report, we process the contact details and information you send only to respond, fix the issue, and keep a record of the report. Do not include secrets, tokens, customer data, or private infrastructure. Security reports go to security@stark-ai.de or GitHub private vulnerability reporting.

This catalog site

The catalog is a static site built from public repository files and hosted on GitHub Pages. We do not run analytics, advertising, or an account system on these pages. GitHub, GitHub Pages, and your browser may process HTTP requests, security logs, or similar technical data under GitHub's notices. We do not set GitHub's retention period, subprocessors, or transfer terms.

The homepage also loads a badge image directly from Product Hunt. Your browser sends Product Hunt the technical data needed to request the image, such as your IP address. The image request omits referrer information. Opening the badge takes you to Product Hunt.