maa-pipeline-generate
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes local Python scripts (
scripts/generate_node.pyandscripts/generate_sweep.py) to perform complex coordinate calculations and handle the merging of generated nodes into existing JSON pipeline files. - [INDIRECT_PROMPT_INJECTION]: The skill's workflow involves reading text from external device screens via OCR, which represents an untrusted data ingestion surface.
- Ingestion points: OCR text results extracted from the
ocr()tool inSKILL.mdandscripts/generate_node.py. - Boundary markers: Not explicitly defined for the OCR content itself, though the skill context strictly limits the use of this data to populating JSON fields.
- Capability inventory: The skill possesses the capability to modify project configuration files (pipeline JSONs) and interact with ADB-connected devices.
- Sanitization: The recognized text is treated as a literal value for matching purposes and is not executed as code, significantly limiting the risk of injection affecting the agent's behavior.
Audit Metadata