The Agent Skills Directory

[COMMAND_EXECUTION]: The skill workflow in SKILL.md recommends using shell commands (find, grep) that incorporate user-provided search terms and vault paths. The pattern find <vault-root> -name "*.md" | grep -i "<search-term>" is susceptible to command injection if the agent does not properly sanitize or escape shell metacharacters provided in the input variables.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted data in a way that could influence agent behavior. Ingestion points: User-provided search terms for vault discovery and external URLs for web cards/iframes are ingested by the agent. Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions embedded in the content discovered during the find command. Capability inventory: The skill encourages the agent to execute shell commands (find, grep) and perform file system writes (creating .canvas JSON files). Sanitization: There are no instructions or best practices provided to the agent regarding the sanitization of user-provided variables before they are interpolated into the recommended shell commands or the generated canvas file structure.

obsidian-canvas