maa-pipeline-guide

Pass

Audited by Gen Agent Trust Hub on Nov 22, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to have the agent ingest and process external data from the user's project environment and provides instructions on how to generate or modify code involving command execution.
  • Ingestion points: The agent is instructed to read project-specific files such as basic_info.md, interface.json, and existing pipeline JSON files in the pipeline/ directory to maintain consistency.
  • Boundary markers: The instructions do not explicitly mandate the use of delimiters or provide warnings to ignore potentially malicious instructions embedded within the processed external data files.
  • Capability inventory: The documentation (field-reference.md) describes the use of the Shell action (for executing arbitrary commands via cmd) and the Command action (for executing binaries via exec). It also details CustomAction and CustomRecognition which involve the execution of Python scripts.
  • Sanitization: There are no specific instructions for sanitizing, escaping, or validating data extracted from untrusted external project files before interpolating that data into generated commands or pipeline logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 22, 2026, 01:27 AM