json-flat-tool

Fail

Audited by Gen Agent Trust Hub on Mar 12, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill's documentation demonstrates piping data from untrusted external URLs (api.example.com) directly into a command string that invokes the python3 interpreter. This pattern is dangerous as it allows untrusted remote content to potentially influence or execute within the command-line context.
  • [DATA_EXFILTRATION]: The jstool.py script includes a feature to read local file content using the '@' symbol followed by a file path. This can be used to read and expose sensitive files such as SSH keys or environment variables by loading them into the JSON data processed by the tool.
  • [EXTERNAL_DOWNLOADS]: The skill documentation encourages fetching and processing data from external, untrusted sources such as api.example.com, which is not a whitelisted or trusted vendor.
  • [PROMPT_INJECTION]: The tool is vulnerable to indirect prompt injection because it processes untrusted data from external sources without proper sanitization or boundary markers.
  • Ingestion points: The jstool.py script reads JSON data from stdin and from specified file paths.
  • Boundary markers: There are no delimiters or specific instructions to ignore embedded commands within the ingested data.
  • Capability inventory: The script can read and write files on the local filesystem using the open() function.
  • Sanitization: Input is parsed with json.loads(), but no content-level validation or filtering is performed.
Recommendations
  • HIGH: Downloads and executes remote code from: https://api.example.com/data - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Mar 12, 2026, 04:50 AM