materialcraft

Pass

Audited by Gen Agent Trust Hub on Nov 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructions direct the agent to discover and execute local binaries (materialcraft and materialcraft-cli) using system utilities such as command -v, where, and mdfind. These commands are used to establish a connection to the tool's automation interface.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external material graph files (.sbs, .mcmat) which contain user-controllable text fields like comments, node labels, and graph descriptions. These fields could be leveraged to provide malicious instructions to the agent during inspection.
  • Ingestion points: Data returned by the session.inspect and library.list methods, as well as file contents read via package.open (SKILL.md and references/control-protocol.md).
  • Boundary markers: The documentation does not specify explicit delimiters or warnings to ignore instructions embedded within graph metadata.
  • Capability inventory: The skill possesses the ability to read and write files via package.save and outputs.export, and to execute various engine commands via command.run.
  • Sanitization: No specific sanitization or filtering logic is described for the text fields extracted from the material graphs.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 13, 2026, 01:50 AM