orienting-codebases

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

• Third-party content exposure detected (high risk: 0.85). The skill fetches and extracts outsider-authored repository content at runtime (e.g., curl ... /tarball/$REF from GitHub, then tar -xzf into /tmp/$REPO), and then feeds extracted free-form code text into the LLM context via the HTML artifact generation step (treesit source extraction → composing-html ... --spec with body_html containing “Actual source extracted by treesit”).