developing-genkit-dart

Pass

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides installation instructions for the Genkit CLI using a script from cli.genkit.dev piped directly to a shell. This is the official installation method for the Genkit ecosystem.- [EXTERNAL_DOWNLOADS]: The documentation references and demonstrates the installation of various packages from public repositories, including the genkit-cli (npm) and the schemantic library (pub.dev).- [COMMAND_EXECUTION]: The skill includes documentation for executing external commands, such as using npx to launch MCP servers and the genkit start command to wrap Dart execution.- [DATA_EXFILTRATION]: Documentation for the FilesystemPlugin and MCP filesystem tools describes capabilities for reading and writing files. These features are intended to enable agents to perform development tasks such as log analysis and code editing within a restricted root directory.- [PROMPT_INJECTION]: The skill describes interfaces that constitute an indirect prompt injection surface.
  • Ingestion points: The FilesystemPlugin and MCP filesystem tools (SKILL.md, references/genkit_mcp.md) allow the agent to ingest content from local files. The SkillsPlugin allows loading system instructions from external files.
  • Boundary markers: The provided code examples do not specify the use of delimiters or instructions to prevent the agent from obeying commands embedded in ingested data.
  • Capability inventory: The framework supports file modification (write_file), network interactions (flow execution via genkit_shelf), and external process execution via MCP.
  • Sanitization: No sanitization or validation of ingested file content is demonstrated in the documentation snippets.
Audit Metadata
Risk Level
SAFE
Analyzed
May 19, 2026, 01:15 PM
Security Audit — agent-trust-hub — developing-genkit-dart