pandaai-workflow-generator
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input through natural language descriptions or strategy_brief.json files to generate Python code for financial strategies. Evidence: Ingestion points are located in the SKILL.md prompt template and scripts/from_qbti_brief.py. Boundary markers are present in the form of a mandatory disclaimer template in references/disclaimer_template.md and explicit agent instructions in SKILL.md to avoid deriving templates by hand once the bridge script runs. Capability inventory: scripts/generate_workflow.py performs local file writes to save generated JSON. Sanitization relies on standard JSON encoding for assembly and the target platform's documented static and runtime sandbox restrictions.
- [COMMAND_EXECUTION]: The validation script scripts/smoke_test.py utilizes subprocess.run to execute local Python scripts during the assembly pipeline verification. This is an expected testing behavior within the local workspace context.
Audit Metadata