coding-harness

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

• Third-party content exposure detected (high risk: 0.80). The skill's required workflow explicitly instructs the agent to fetch and interpret live, repo- and web-hosted content (e.g., harness verify-coderabbit / GitHub API calls, the curl GET to the CircleCI API shown in references/agent-install-guide.md and references/agent-install.json, npm view, and reading package.json/.codex/environments/environment.toml and .coderabbit.yaml) and to base validation and migration decisions on those responses, so untrusted third-party or user-generated content can materially influence next actions.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly requires installing the private npm package @brainwav/coding-harness from the npm registry (e.g., mise install -g npm:@brainwav/coding-harness which resolves to https://registry.npmjs.org/), which fetches remote code that will be installed and executed by the harness CLI at runtime.