knowledge-impact
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes shell commands including
rg(ripgrep) for searching knowledge catalogs andsedfor inspecting specific lines within repository files. - [COMMAND_EXECUTION]: The skill executes a local script located at
~/projects/llmops-wiki/catalog/scripts/lint-catalog.shto validate changes made to the catalog files. - [INDIRECT_PROMPT_INJECTION]: The skill operates on a surface vulnerable to indirect prompt injection by ingesting and processing data from external project files.
- Ingestion points: Markdown files within
~/projects/llmops-wiki/catalog/views/and~/projects/llmops-wiki/silver/entity-alias/index.md. - Boundary markers: Absent; there are no specific instructions or delimiters to distinguish between data and instructions within the catalog files.
- Capability inventory: Shell command execution (
rg,sed,bash) and the ability to append confirmed findings back to the catalog files. - Sanitization: No evidence of sanitization, escaping, or validation of the content retrieved from the catalogs before it is processed by the agent.
Audit Metadata