command-development

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

• Third-party content exposure detected (high risk: 0.80). This skill includes commands that fetch and ingest untrusted, user-generated content from third-party sources—e.g., gh pr view $1 --json, gh pr diff $1 --name-only, and other git/gh/bash calls that read GitHub PRs/diffs and repository files—which the agent is expected to read and interpret as part of its review/workflow.