linkml-author

Pass

Audited by Gen Agent Trust Hub on Oct 23, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data sources (source code, database DDL, OpenAPI/GraphQL docs) to derive schema elements, which could allow malicious instructions in those files to influence the agent. Ingestion points: Reads from project directories like 'src/' and 'models/'. Boundary markers: No explicit delimiters or instructions to ignore embedded commands are specified. Capability inventory: Can execute shell commands via 'just' and 'uvx', and write to schema YAML files. Sanitization: No explicit content filtering is mentioned.
  • [COMMAND_EXECUTION]: The skill instructs the agent to run various shell commands for linting, generation, and testing using task runners like 'just' and 'make', and Python tools like 'uv' and 'poetry'. It also executes a project-specific preflight script 'linkml-preflight.sh'.
  • [DYNAMIC_EXECUTION]: The skill generates Python dataclasses and Pydantic models from schemas using tools like 'gen-python' and 'gen-pydantic', and subsequently executes test suites that utilize this generated code.
  • [EXTERNAL_DOWNLOADS]: The skill uses 'uvx' to fetch and run the 'linkml' and 'rust-just' toolsets from the public PyPI registry when they are not already installed in the local environment, which is a standard ecosystem practice.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 23, 2026, 07:57 PM