session-loop

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

• Third-party content exposure detected (medium risk: 0.30). In sub/catchup.md the runtime reads free-form project handoff content from NEXT_SESSION.md and UPDATES.md (top entry) and the last 3 ledger ADs, then uses their extracted text to drive drift-check behavior and resume recommendations.

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

• Potentially malicious external URL detected (high risk: 0.90). The README explicitly instructs cloning and running remote code via "git clone https://github.com/akshayrao14/session-loop" followed by running the repo's install.sh, which fetches and executes external code that determines the skill's runtime behavior.