extended-thinking-architect
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security concerns were identified during the analysis of this skill. The instructions and scripts focus on cost optimization and reasoning strategy for AI agents.
- [COMMAND_EXECUTION]: The skill references local execution of two Python scripts (
reasoning_budget_advisor.pyandreasoning_loop_allocator.py). Both scripts use only the Python standard library and perform deterministic calculations without network access or sensitive file operations. - [INDIRECT_PROMPT_INJECTION]: The skill has a minimal attack surface for indirect injection.
- Ingestion points: Command-line arguments passed to the script tools in
SKILL.md. - Boundary markers: Not applicable as the scripts do not interpolate data into LLM prompts within the skill code.
- Capability inventory: The scripts perform no subprocess calls, file writes, or network operations.
- Sanitization: Standard argument parsing is used to validate inputs against defined choices.
Audit Metadata