wiki-query
Pass
Audited by Gen Agent Trust Hub on Nov 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions provide shell command templates for tools such as
obsidian-wiki,qmd, andgrepthat directly interpolate user-supplied query terms and path variables (e.g.,obsidian-wiki graph-query "$OBSIDIAN_VAULT_PATH" "<question>"andGrep -A 10 -B 2 "<query-term>" <candidate-file>). This pattern poses a risk of command injection if the underlying execution environment or the agent does not properly sanitize these inputs. - [INDIRECT_PROMPT_INJECTION]: The skill processes data from an external Obsidian vault, creating a surface for indirect prompt injection where instructions embedded in wiki pages could influence agent behavior.
- Ingestion points: The skill reads
index.md,hot.md,.manifest.json, and various markdown files (.md) throughout the vault during its retrieval phases. - Boundary markers: The skill includes explicit "READ-ONLY" directives and instructions to refuse user requests for vault modifications, instead suggesting alternative skills for those actions.
- Capability inventory: The skill can execute CLI tools (
obsidian-wiki,qmd,grep), perform extensive file reads within the vault path, and append to a log file via a dedicated utility. - Sanitization: No specific filtering or sanitization process is described for the content retrieved from wiki pages before it is synthesized and presented to the user.
Audit Metadata