writing-agents

Pass

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The recommended workflow for creating agents involves using web search to gather best practices and anti-patterns, which are then incorporated into the agent's system instructions.
  • Ingestion points: External data from web search results is ingested in SKILL.md during the 'Research Domain Best Practices' phase.
  • Boundary markers: No specific delimiters or instructions to ignore commands within search results are mandated for the agent creation process.
  • Capability inventory: The skill facilitates the creation of agents that execute shell commands and file system operations via the Task tool.
  • Sanitization: There is no requirement or guidance for sanitizing external content before it is used to define an agent's persona and rules.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 28, 2026, 10:13 AM