santa-method
Pass
Audited by Gen Agent Trust Hub on Apr 6, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The
REVIEWER_PROMPTinSKILL.mdinterpolates the{output}variable (content generated by a previous agent) directly into the reviewer agent's context using only Markdown headers as delimiters.\n - Ingestion points: The
{output}variable inSKILL.mdreceives data from a generator agent.\n - Boundary markers: The prompt uses Markdown headers (e.g.,
## Output Under Review) to separate the data from instructions.\n - Capability inventory: The skill uses the
Agenttool to spawn subagents and afix_agentto modify content based on reviews.\n - Sanitization: There is no evidence of escaping, validation, or filtering of the interpolated
{output}content before it is passed to the reviewer agents.
Audit Metadata