maa-pipeline-option
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the use of shell commands for local validation of project resources and JSON syntax.
- Evidence:
python tools/ci/check_resource.py assets/resource/baseandrg "get_node_data|_node_enabled|Flag_" agent assetsin SKILL.md. - [INDIRECT_PROMPT_INJECTION]: The skill defines a protocol for 'input' options that allow external data to be injected into the pipeline's execution context at runtime.
- Ingestion points: User-provided values via the
inputtype defined inassets/interface.json(documented in SKILL.md and references/protocol.md). - Boundary markers: The system relies on structured JSON fields (
expected,custom_action_param) to isolate input, but lacks explicit LLM-specific boundary delimiters. - Capability inventory: Python logic (e.g.,
context.get_node_data) retrieves these values to influence recognition parameters and action execution flow. - Sanitization: The protocol includes a
verifyfield for regex-based validation to ensure input conforms to expected patterns (documented in references/protocol.md).
Audit Metadata