semble-setup

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 1.00). The only suspicious item is a direct link to a shell installer (https://astral.sh/uv/install.sh), which is a high-risk pattern because running remote .sh scripts from an untrusted domain can deliver malware or run arbitrary commands.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). SKILL.md only routes/installs/repairs; at runtime the hooks it wires (assets/semble-prefetch.mjs on UserPromptSubmit and assets/semble-reminder.mjs on PreToolUse Bash|Grep) consume free-form user prompt text and derive search queries, which are then used to query the MCP-backed code index.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs the agent to run shell scripts that install packages (brew/uvx), write and modify user- and repo-level configuration and hook files (e.g. ~/.claude.json, .claude/hooks/, CLAUDE.md, .sembleignore), create/reserve cache roots, and delete cache directories (purge), so it clearly mutates the machine state outside the project.