agent-identity-evolution

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

• Third-party content exposure detected (high risk: 0.85). The required runtime workflow includes /evolve sync and /evolve discover, which import and process text from connected outsider-authored sources (e.g., public websites, RSS feeds, GitHub repos, Twitter/X, LinkedIn exports) into the platform’s PBD pipeline and then into the agent’s evolved system prompt/LLM context via /evolve prompt.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill makes runtime API calls (e.g., curl "$LIVE_NEON_BASE/agents/$AGENT_ID" and "$LIVE_NEON_BASE/agents/$AGENT_ID/resolved-identity" — i.e. https://agentpersona.live/api/v1/agents/{agentId}) to fetch a ".system_prompt" which is then injected as the agent's system instruction, so remote content directly controls the agent prompt at runtime.