maa-project-init
Pass
Audited by Gen Agent Trust Hub on Nov 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill scans and summarizes content from untrusted local files (such as interface.json, pipeline JSON files, and Python scripts) into a basic_info.md document. This document is intended to be used as context for the agent in subsequent tasks. If these project files contain malicious instructions disguised as data (e.g., in node names, task descriptions, or strings), the agent might inadvertently follow them when reading the summary.
- Ingestion points: The
scripts/analyze_pipeline_project.pyscript reads multiple files from the target project root, including JSON configurations and Python source code. - Boundary markers: The script uses Markdown tables and Mermaid code blocks to separate data fields, providing some structural separation between different pieces of ingested data.
- Capability inventory: The script performs file reads and writes (specifically creating
basic_info.md). It does not perform network operations. - Sanitization: The script includes
md_escapeandmermaid_labelfunctions to escape characters that might break Markdown or Mermaid syntax, which provides a layer of defense against structural manipulation of the output document. - [COMMAND_EXECUTION]: The skill instructions in
SKILL.mddirect the agent to execute a local Python script (analyze_pipeline_project.py) via the command line to perform the project scan. While this is the intended functionality, users should ensure they are running the skill against trusted project repositories to avoid unexpected script behavior.
Audit Metadata