wiki-workflow

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

• Third-party content exposure detected (high risk: 0.90). This skill's INGEST workflow explicitly fetches arbitrary URLs ("If it's a URL, fetch it." in SKILL.md — INGEST step 1) and saves/extracts their rendered markdown into /code/wiki/sources/ which the agent then reads and acts on to plan and create/update pages, exposing it to untrusted third‑party content that could carry injected instructions.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly states "If it's a URL, fetch it" and saves/uses the fetched content as source material that is injected into the agent's processing to synthesize and write wiki pages, so arbitrary user-provided HTTP/HTTPS URLs (e.g., any http(s)://... URL supplied at ingest time) can directly control the agent's prompts/model context.