prompt-structurer
Pass
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted user prompts, creating an indirect prompt injection surface where embedded instructions could attempt to override the structuring process.\n
- Ingestion points: Untrusted content enters via
PROMPT_TEXT,SUITE_CONTEXT, andEXISTING_XML_PROMPTas defined inSKILL.mdand processed by the subagent registry.\n - Boundary markers: The skill uses explicit tags to wrap user data and includes specific instructions for the agent to treat these blocks as inert text to analyze.\n
- Capability inventory: The skill has capabilities for file-system writes (to a user-specified
OUTPUT_TARGET) and limited network fetching.\n - Sanitization: Security depends on the LLM adhering to the provided boundary instructions and inert-text directives.\n- [EXTERNAL_DOWNLOADS]: The skill maintains a fetch policy for external rationale when local references are insufficient.\n
- Source Evaluation: Authorized sources include official documentation from established organizations and technical guidance from recognized community repositories.\n
- Mitigation: The fetch budget is strictly limited to one URL per run, managed exclusively by the orchestrator.
Audit Metadata