tokhub

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

• Third-party content exposure detected (medium risk: 0.65). High indirect prompt-injection risk: the required runtime workflow can execute scripts/fetch-docs.sh, which performs a network curl to fetch README/CONTRIBUTING/API/etc. from GitHub and then those fetched (outsider-authored) free-text docs could be ingested into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs re-fetching live docs at runtime via scripts/fetch-docs.sh which curls https://raw.githubusercontent.com/${OWNER_REPO}/main/... to pull remote README/CONTRIBUTING content that would be incorporated into guidance, so the raw.githubusercontent URL is a runtime external dependency that can control prompts.