validate-yaml
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFEDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [DYNAMIC_EXECUTION]: The skill documents the use of the
allow_python=Trueparameter in methods likeyaml_interrogate()andyaml_ablation()to enable the execution of Python code blocks defined within YAML files. - Evidence:
SKILL.mdandreferences/yaml-schema.mddescribepython:,pre:,expr:, andexpressions:keys that execute strings as Python code (e.g.,lambdafunctions or module-level functions). - The skill explicitly warns that this code "can do anything the Python session can" and advises only allowing it for trusted configurations.
- [COMMAND_EXECUTION]: The skill introduces the
pb runcommand-line interface, which allows users to execute validation plans directly from the terminal, including plans with embedded Python code. - [INDIRECT_PROMPT_INJECTION]: The skill creates a surface for indirect injection by processing external YAML files that may contain malicious instructions or code.
- Ingestion points: Data is ingested via
pb.yaml_interrogate(),pb.yaml_ablation(), and thepb runCLI command using external YAML files (e.g.,validation.yaml). - Boundary markers: The skill relies on the
allow_pythonflag as a security boundary, which isFalseby default. - Capability inventory: When enabled, the skill allows arbitrary Python execution, including file access and network operations (as defined by the Python environment).
- Sanitization: The underlying library raises
YAMLValidationErrorfor unauthorized code blocks if the safety flag is not explicitly set. - [CREDENTIALS_UNSAFE]: The documentation provides examples of database connection strings that include placeholders for user credentials in plaintext.
- Evidence:
references/yaml-schema.mdcontains the exampletbl: "postgresql://user:pass@host:5432/db::orders".
Audit Metadata