skillify

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

• Third-party content exposure detected (medium risk: 0.30). In CREATE mode, the workflow explicitly reads backward through “this conversation” to extract repeated workflow details (from-chat Step 4, which ingests user-authored free text at runtime).

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's ADAPT flow explicitly runs git/gh fetches at runtime (e.g., "git clone --depth 1 " and "gh api ... repos///contents/SKILL.md") which pulls external SKILL.md content from https://github.com// that will be parsed and used to control the adapted skill's instructions.