night-market-collective-memory

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md indicates the required workflow reads GitHub Discussions at runtime via gh api graphql queries (e.g., fetching recent discussions, and optionally fetching a discussion body and comments), which ingests outsider-authored free text from users posting to those discussions without any demonstrated pre-selection/sanitization barrier.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly calls "gh api graphql" (e.g., gh api graphql -f query=...) to fetch GitHub Discussions (repo:athola/claude-night-market) at runtime and injects those discussion bodies as the agent's collective memory, meaning remote GitHub content directly influences agent prompts and behavior.