twitter-api

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s required runtime workflow can ingest outsider-authored free text from X by calling endpoints like GET https://twitter.fetcher.sh/api/search with a user-supplied query (e.g., operators/keywords/handles) that returns matching tweet text at runtime.

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

• Potentially malicious external URL detected (high risk: 0.80). I flagged this because the skill exposes an MCP endpoint (https://twitter.fetcher.sh/mcp) and machine-readable artifacts (https://twitter.fetcher.sh/openapi.json, https://twitter.fetcher.sh/llms.txt, https://twitter.fetcher.sh/skill.md) that an MCP-connected agent can fetch at runtime to discover/define tools and thus directly influence agent prompts/behavior.