ai-repo-safety

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.90). While many links are legitimate (PyPI, known GitHub/GitLab projects, npm/registry pages), the set includes high-risk items — a plainly malicious domain (https://evil.com/shell.sh), direct raw install scripts (raw.githubusercontent.com/*/install.sh / install.ps1) and direct-release downloads (GitHub release .tar.gz) which, if executed without verification, make this a suspicious download source.

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

• Third-party content exposure detected (high risk: 0.85). Outsider free text can enter the LLM context via the github-guard read / gitlab-guard read runtime path, which fetches issue/PR/commit bodies from external GitHub/GitLab APIs (gh api / glab api) and prints the sanitized payload as JSON for the agent to consume.