report-limit

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

• Third-party content exposure detected (medium risk: 0.30). The required workflow reads outsider-authored free text from the user’s own arguments/context (e.g., the date argument) and uses it to fetch/categorize cached timeline windows, meaning attacker-controlled free text can enter runtime logic; there is also external page fetching for pricing only as first-party documentation.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly performs a runtime WebFetch of https://platform.claude.com/docs/en/about-claude/pricing#model-pricing to parse authoritative model fields and then edit local model-pricing.json and re-run the script, so fetched content directly controls subsequent agent behavior.