sendcutsend

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

• Third-party content exposure detected (high risk: 0.75). The required workflow downloads outsider-authored public web content at runtime (SendCutSend ordering guide/catalog/specs from cdn.sendcutsend.com) via python scripts/download_sources.py, then the skill instructs the agent to read the generated files in references/generated/ and use them as evidence for report writing—so readable markdown/JSON from the web enters the LLM context indirectly.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill's runtime downloader (scripts/download_sources.py) fetches live rule/data files such as https://cdn.sendcutsend.com/specs/sendcutsend-ordering-guide.md which are used as the authoritative rule sources that directly drive the agent's validation logic and report decisions.