teach-me

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime LLM context is populated by the chapter-writing agents’ prompts, which include Phase 4 research results fetched from public web pages (Lane A WebSearch+WebFetch), so outsider-authored free text can be ingested into the LLM context via those research reports.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill fetches and executes remote JavaScript at build/runtime: it explicitly WebFetches https://cdn.jsdelivr.net/npm/marked/marked.min.js and runs npm --prefix <skill-dir>/assets install which pulls packages from the npm registry (e.g. https://registry.npmjs.org/) to install and run Shiki — both are runtime downloads of remote code that will be executed as part of the build/viewer.