evolve

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

• Third-party content exposure detected (high risk: 1.00). The repository explicitly ingests public third‑party content (e.g., scripts/sync-skills.sh that syncs skill metadata from GitHub and numerous "source:" links to X/Threads/GitHub in .claude/memory/), and the skill workflow (CP1 memory_search, Memory MCP integration in skills/05-integration/_base/ and the north-star sync scripts) requires the agent to read and act on those indexed external items, so untrusted user-generated content can materially influence tool use and decisions.

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

• Potentially malicious external URL detected (high risk: 0.80). The repo content from GitHub (e.g., https://github.com/ZhangHanDong/makepad-skills) is referenced as an installable skill/source and the doc explains skillpkg installs/merges remote SKILL.md into the installed SKILL.md which is then loaded into the agent runtime context—meaning fetched repo content can directly control prompts/instructions at runtime.