craftapps

Pass

Audited by Gen Agent Trust Hub on Nov 1, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill core functionality relies on the execution of CLI tools (e.g., deckcraft-cli, wordcraft-cli, pdfcraft-cli) and the operation of MCP servers. It specifically instructs the agent to use network connections to local ports (127.0.0.1:PORT) for 'bridge mode' to interface with running GUI applications.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from untrusted external sources through complex file formats, which provides a surface for indirect prompt injection.
  • Ingestion points: Untrusted data enters the agent context via files (PDF, DOCX, XLSX, SVG, etc.) processed by the apps listed in SKILL.md and routing.md.
  • Boundary markers: The instructions lack boundary markers or warnings to the agent to disregard instructions embedded within the processed file content.
  • Capability inventory: The agent possesses CLI execution capabilities used to read and write files and establish local network bridge connections.
  • Sanitization: No sanitization, validation, or filtering of extracted content is described before it is introduced to the agent's context.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 1, 2026, 06:08 PM