security-osint

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs the agent to ask users to paste API keys or OAuth codes into the conversation and then shows code/commands that embed those secrets verbatim (e.g., "THE_KEY_FROM_USER", "THE_CODE_FROM_USER", Authorization header), which requires the LLM to handle/output secret values directly.

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime workflow calls Xpoz tools to fetch Twitter/X and Reddit posts (including user-authored text/title fields) via getTwitterPostsByKeywords, getRedditPostsByKeywords, and the SDK twitter.search_posts/reddit.search_posts, so outsider-authored free text can be ingested before analysis.