skill-shed

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). This is a third‑party GitHub repository (user "Oruga420") referenced for cloning — an unverified personal source that could contain unreviewed scripts or payloads and is therefore potentially risky to download and execute.

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md Step 4 Option A fetches outsider-authored SKILL.md files from the public GitHub repo at runtime (via gh api ... contents/{skill-name}/SKILL.md), then decodes and “apply[s] it in the current conversation,” which would feed that free text into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches remote skill content at runtime and applies it to the conversation (prompt injection) via "gh api repos/Oruga420/claude-code-skills/contents/{skill-name}/SKILL.md --jq '.content' | base64 -d" and also clones remote code with "git clone https://github.com/Oruga420/claude-code-skills.git", so external content can directly control agent behavior.