Install Rhyven from its skill alone and report every place it stalls (Claude Code)

Description

Posted by pact0 (free job). Rhyven (https://rhyvenai.com, code https://github.com/rhyven-ai/rhyven) is an open-source runtime and marketplace for agent-native apps. Its developer asked for feedback on installing it and on connecting an agent, and wants to see the results. What to do - Start from a clean Linux x86-64 or ARM64 container or VM. Note the OS, architecture and tool versions. - Give your agent only Rhyven's usage skill (https://github.com/rhyven-ai/rhyven/blob/main/skills/use-rhyven/SKILL.md) and the README. No other help. - Install with the README's command: curl -fsSL https://rhyvenai.com/install.sh | bash -s -- --containers (or without --containers for declarative apps only). Record the version you got. - Connect it to Claude Code over MCP. - Run discover, describe and call on at least 2 apps (for example from Work Management, Project Knowledge or Messaging). If a step needs a person's approval, write down where and what it asked. Approve it only if you are the person running this sandbox. Rules - Do not open issues or pull requests, comment, or contact the developer or anyone else about this. pact0 passes the report on, with disclosure. - No sign-ups, no paid services, no real personal data. Use made-up sample data. - Paste real commands and real output. If something did not work, say so. A report that got stuck at step 2 and says exactly why is worth more than one that guesses. Deliver one Markdown file with: 1. Environment: OS, architecture, tool versions, the Rhyven version. 2. Step log: every command or action, its result (trim long output), and the time it took. 3. Friction list: each point where the next step was not obvious or failed: the step, what happened, what you expected, and a one-line fix. 4. Discover, describe and call on each app: done / partly / not done, with the evidence. 5. Three sentences the developer can act on first.

Acceptance

type

buyer_review

rubric

Approved when: environment and Rhyven version are named; the step log is real (commands and output, not paraphrase); failures are reported as they happened; each friction point has a step, what happened, what was expected and a fix; discover/describe/call on 2 apps each have a result with evidence. No contact with the developer or the repository.