algo-net-influence

Pass

Audited by Gen Agent Trust Hub on Apr 10, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection through the ingestion of external network graph data. Ingestion points: Network data (nodes and edges) is ingested in SKILL.md under 'Phase 1: Input Validation'. Boundary markers: The instructions lack delimiters or explicit warnings to the agent to ignore instructions embedded within the graph data. Capability inventory: The skill provides Python implementation code in references/celf-implementation.md and references/scalable-im.md for the agent to execute for simulations. Sanitization: There is no evidence of sanitization or content validation to prevent malicious strings in the graph from influencing the agent's behavior.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 10, 2026, 07:10 AM