connection-auth-rules

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

• Third-party content exposure detected (high risk: 1.00). The skill's fetch_schema.py explicitly fetches and parses live connector code and docstrings from the public GitHub repo (via https://api.github.com/repos/monte-carlo-data/apollo-agent/contents), and the SKILL.md requires the agent to read that untrusted, user-editable content to drive mapper/step construction, so external content can materially influence actions.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill's fetch_schema.py makes runtime requests to the GitHub API (https://api.github.com/repos/monte-carlo-data/apollo-agent/contents) and the returned download_url/raw GitHub file URLs and uses the fetched Python source and docstrings to drive the connector schema and transform-step information presented to the user, so external content directly controls the skill's prompts/behavior.