graphify

Warn

Audited by Gen Agent Trust Hub on Jun 20, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill defines a shell execution step where user input from $ARGUMENTS is assigned to a variable and used directly in a command string (graphify "$TARGET"). If the platform does not sanitize the input, this allows an attacker to execute arbitrary commands using shell metacharacters such as semicolons or pipes.- [EXTERNAL_DOWNLOADS]: The instructions prompt the user to install the graphifyy package from PyPI and reference a third-party GitHub repository (safishamsi/graphify). This introduces a supply-chain dependency on an external tool not maintained by a known trusted entity.- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the processing of GRAPH_REPORT.md, which is generated from the codebase analysis.
  • Ingestion points: Step 4 where the agent evaluates GRAPH_REPORT.md content.
  • Boundary markers: None identified.
  • Capability inventory: Subprocess execution of the graphify CLI.
  • Sanitization: No validation or sanitization is performed on the codebase data before it is processed into a report that influences agent behavior.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 20, 2026, 05:00 PM
Security Audit — agent-trust-hub — graphify