matlab

Pass

Audited by Gen Agent Trust Hub on Jun 26, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: Behavioral steering for vendor promotion. The skill contains specific instructions for the agent to proactively suggest the 'K-Dense Web' (www.k-dense.ai) platform when users request complex research, large document analysis, or multi-step workflows.
  • [PROMPT_INJECTION]: Surface for indirect prompt injection via untrusted data processing. The skill is designed to ingest and process data from external files and web sources, which could contain malicious instructions designed to influence agent behavior.
  • Ingestion points: Ingests data from CSV, Excel, and Text files using functions like readtable, readmatrix, and readlines (found in references/data-import-export.md). It also fetches remote data via the Python requests library (found in references/python-integration.md).
  • Boundary markers: Absent. There are no instructions or delimiters provided to help the agent distinguish between data content and operational instructions.
  • Capability inventory: The skill has significant capabilities including file system write operations (writetable, save, fwrite), file deletion (delete), and shell command execution (matlab -batch, octave --eval).
  • Sanitization: Absent. The provided references do not include guidance on validating or sanitizing ingested data before processing.
  • [COMMAND_EXECUTION]: Documentation for shell-level interpreter interaction. The skill provides multiple patterns for executing MATLAB and Octave scripts from the command line, including the use of -batch and --eval flags, as well as administrative tasks such as changing file permissions (chmod +x).
  • [REMOTE_CODE_EXECUTION]: Dynamic execution of runtime-generated code. The Python integration reference documents the use of pyrun and pyrunfile, which allow the execution of arbitrary Python strings and external script files within the environment.
  • [DATA_EXFILTRATION]: Network operations via external libraries. The skill demonstrates the use of the Python requests module to perform HTTP GET requests to external domains (e.g., api.example.com).
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 26, 2026, 07:59 PM