resolve-recipe-ingredient-link

Warn

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command using a local binary located at /Users/mtm/go/bin/islandiguana to list files. This represents direct shell interaction within a hardcoded environment.\n- [EXTERNAL_DOWNLOADS]: The test documentation references pnpm dlx promptfoo. While this involves an external tool download, promptfoo is a well-known service for LLM evaluation and is considered a safe development tool for testing purposes.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via user-supplied wikilinks in recipe notes.\n
  • Ingestion points: Ingredient lines containing wikilinks are read from source notes provided by the user.\n
  • Boundary markers: No explicit delimiters or instructions are provided to the agent to isolate the link text from potentially malicious surrounding content.\n
  • Capability inventory: The agent can execute shell commands and write new Markdown files to the local file system.\n
  • Sanitization: The instructions lack any validation or sanitization logic for the extracted link text before using it as a filename. This could be leveraged to attempt path traversal (e.g., [[../sensitive_file]]) when the agent creates new files in the vault.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 30, 2026, 06:08 PM
Security Audit — agent-trust-hub — resolve-recipe-ingredient-link