writing-technical-design

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

• Third-party content exposure detected (high risk: 0.90). The skill's Deep Research Phase (SKILL.md and references/DEEP-RESEARCH.md) explicitly requires using WebSearch/WebFetch and spawning research subagents to fetch and summarize public web pages, official docs, blog posts, and forum content into component SKILL.md files that are then auto-loaded and used to drive design decisions and agent behavior, so untrusted third‑party content is ingested and can materially influence actions.

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

• Potentially malicious external URL detected (high risk: 0.70). This skill explicitly instructs runtime WebFetch/WebSearch usage to pull authoritative pages (e.g., https://redis.io/docs/latest/develop/data-types/streams/) as part of its mandatory "Deep Research" phase and to inject summarized content into per-component SKILL.md files that are auto-loaded into agent context, meaning external pages fetched at runtime can directly influence agent prompts/instructions.