pp-judgeme

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The skill README describes an unconditional local "learning"/teach loop that journals every invocation (including examples showing live "token" values), combined with webhook delivery and feedback auto-send features — creating clear vectors for accidental or intentional data exfiltration and credential leakage if misused or misconfigured.

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow uses the user-provided question text in recall and then reads the CLI’s returned candidates, playbook.notes, notes, results, and warnings to drive subsequent command execution; this means outsider-authored free text (the user prompt) is ingested to select which commands/tools to run.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill's prerequisite/install steps instruct using go install on GitHub module paths (github.com/mvanhorn/printing-press-library/library/marketing/judgeme/cmd/judgeme-pp-cli@latest and github.com/mvanhorn/printing-press-library/library/marketing/judgeme/cmd/judgeme-pp-mcp@latest), which fetch and build/execute remote code during setup and are required for the CLI the skill relies on.