paperclip-plan-work

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

• Third-party content exposure detected (medium risk: 0.30). In paperclip-plan-work, the runtime reads the selected Paperclip parent issue/parent chain, its plan document, blockers, and comments (step 2) to propose child issues, so outsider-authored free text in those comments/descriptions can be ingested without pre-selecting a specific item.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The docs instruct running remote package installs/execs such as "npx -y @bbengamin/paperclip-mcp-server" and "npm install -g paperclipai", which fetch and execute remote code at runtime and are referenced as required/fallback dependencies for the skill (e.g., MCP server), so they present a runtime external dependency that executes remote code.