dag-library

Warn

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill utilizes dynamic ESM imports in JavaScript where the module path is constructed at runtime using an environment variable (OMO_DAG_SDK_ROOT). It also dynamically constructs task definitions from local files that are subsequently parsed and executed.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests DAG definitions from JSON files stored in directories such as $PWD/.omo/dags. This creates an attack surface where malicious instructions could be introduced via untrusted files included in a cloned repository.
  • Ingestion points: SKILL.md describes loading definition files from $OMO_DAG_LIBRARY, $PWD/.omo/dags, and $HOME/.omo/dags.
  • Boundary markers: No boundary markers or specific delimiters are mentioned to separate instructions from data within the processed JSON prompts.
  • Capability inventory: The skill is capable of writing JSON files to the filesystem and executing multi-node workflows through the tool.dag interface.
  • Sanitization: Basic validation is performed by rejecting filenames containing path characters to prevent simple traversal during save operations.
  • [COMMAND_EXECUTION]: The skill facilitates the execution of complex task sequences (DAGs) using prompts and logic read directly from external JSON files on the disk.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 8, 2026, 05:14 AM
Security Audit — agent-trust-hub — dag-library