omh-codebase-uml
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes local shell commands to perform repository analysis and render images.
- Evidence: Executes
omh codegraph umlfor diagram generation andgit ls-filesto count source files. - Evidence: Mentions invoking
javaorPLANTUML_JARfor the rendering step. - [INDIRECT_PROMPT_INJECTION]: The skill analyzes external, untrusted codebase data, creating a surface for indirect prompt injection via malicious symbol names or file paths.
- Ingestion points: Reads Python files and repository metadata (file names and public symbols) from the local file system (SKILL.md).
- Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard malicious instructions embedded in the symbol names or codebase structure.
- Capability inventory: The agent has capabilities to execute shell commands (
omh,git,java) and read the file system. - Sanitization: No sanitization or validation steps are described for the public symbols and unit names extracted from the code before they are used in the diagram generation or read back to the user in the legend.
Audit Metadata