kb-init
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local shell commands to detect environmental tools and retrieve vault metadata. It specifically runs 'which obsidian' to find the executable, 'obsidian version' for compatibility checks, and 'obsidian tags' to extract established vault conventions.
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted data from existing vault files to generate summaries and update the AGENTS.md configuration.
- Ingestion points: Reads content from user-provided overview files (e.g., *-Project.md, project-overview.md) and local templates.
- Boundary markers: No delimiters or specific 'ignore instructions' markers are mentioned for the content being interpolated into indexes.
- Capability inventory: Includes file system read/write access and the ability to execute the Obsidian CLI tool.
- Sanitization: The skill performs template placeholder substitution without explicit sanitization of the content extracted from the overview files.
Audit Metadata