searching-sourcegraph

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md explicitly instructs the agent to search and read public GitHub repos (e.g., repo:^github.com/ORG/REPO$, repo:github.com/myorg/) and to use tools like sg_read_file and sg_deepsearch_read to ingest and interpret code from those sources, so untrusted, user-generated repository content could influence decisions and actions.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill issues Sourcegraph queries that fetch files from GitHub at runtime (e.g., repo:^github.com/ORG/REPO$ used with sg_read_file / sg_deepsearch_read), meaning external repository content is retrieved and injected into the agent's context and can directly influence prompts/responses.