component-gallery

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

• Third-party content exposure detected (high risk: 0.90). The skill ingests and queries public, user-authored web content (scraped from component.gallery via the crawl_site/firecrawl pipeline in scripts/ingest.py and additional deep-dive markdown fetched from raw GitHub URLs in scripts/ingest.py), and the RAG/query workflow (SKILL.md and scripts/query.py) has the agent read and synthesize that untrusted third‑party content to drive design decisions, which could allow indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.90). The ingest pipeline fetches external markdown at runtime (e.g., https://raw.githubusercontent.com/inbn/component-gallery/main/src/content/componentContent/{filename} and also crawls https://component.gallery/) and uses that fetched content to build the RLAMA RAG collection that will be retrieved into the model context, so remote files can directly control prompts/agent outputs.