agent-communication

Pass

Audited by Gen Agent Trust Hub on Nov 12, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill documents patterns where agents process data from shared communication channels, creating a standard surface for indirect prompt injection.
  • Ingestion points: Data is ingested from shared state in references/patterns.md via blackboard.query and blackboard.read methods.
  • Boundary markers: The BlackboardAgent implementation in references/patterns.md uses simple newline and label delimiters to separate untrusted external context from the system prompt instructions.
  • Capability inventory: The provided scripts do not contain dangerous capabilities such as filesystem writes, subprocess execution, or arbitrary network operations.
  • Sanitization: While the skill emphasizes schema validation (using Zod) for data structure integrity, it lacks specific sanitization or escaping of the string content before interpolation into LLM prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 12, 2026, 07:08 PM