Prose

OpenProse VM skill pack. Activate on any `prose` command, .prose files, or OpenProse mentions; orchestrates multi-agent workflows.

This skill is internally consistent: it is a runner/orchestrator for .prose programs and its instructions match that purpose. Main risks to consider before installing or running it: (1) it fetches and executes .prose programs from arbitrary URLs or registry shorthands — do not run programs from untrusted sources, as they can perform network calls, spawn agents, or read project files; (2) it uses project-level config (e.g., .prose/.env and .prose/agents/) — avoid storing secrets there unless you trust the skill and the code you run; (3) if you want to use external state backends (Postgres/SQLite), expect to provide connection details at runtime — supply only credentials you trust and prefer ephemeral tokens. If you need to be extra cautious, review any remote .prose content before executing and keep secrets out of project config files.

Type: OpenClaw Skill Name: prose Version: 0.1.0 This skill bundle is classified as suspicious due to several high-risk capabilities, despite the stated intent to build a functional AI programming language and VM. Key indicators include instructions for the agent to use `exec` with `curl` for arbitrary command execution (found in `SKILL.md` and `prose.md`), the ability to fetch and execute `.prose` programs from arbitrary remote URLs (`SKILL.md`, `prose.md`), and the explicit handling and exposure of database credentials (`OPENPROSE_POSTGRES_URL`) to subagents and logs (detailed in `SKILL.md` and `state/postgres.md`). While the documentation includes warnings and best practices, these capabilities introduce significant supply chain risks and potential for credential compromise or unauthorized execution if misused or exploited.