nexus-proposal-scout

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

• Third-party content exposure detected (high risk: 0.85). The workflow fetches PHASE.md from repos/<org.github_org>/.github/contents/... via gh api ... --jq .content | base64 -d, which ingests outsider-authored free text from a GitHub org’s repository into the agent’s LLM context at runtime.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches the org's PHASE.md at runtime using the GitHub API endpoint (gh api repos/<org.github_org>/.github/contents/profile/PHASE.md --jq .content | base64 -d) and requires that content to determine what to propose, so remote content directly controls the agent's instructions.