pentest-ctf-binary

Warn

Audited by Gen Agent Trust Hub on Feb 18, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • COMMAND_EXECUTION (MEDIUM): The execute_command tool in references/tools.md allows the agent to run arbitrary shell commands on the host. This is a powerful primitive that could be misused if the agent's behavior is influenced by malicious input.
  • REMOTE_CODE_EXECUTION (MEDIUM): Several tools, including pwntools_exploit, ghidra_analysis, gdb_analyze, and angr_symbolic_execution, allow the execution of dynamic scripts. These allow for arbitrary code execution within the respective tool environments.
  • DATA_EXFILTRATION (LOW): The skill provides tools like strings_extract, xxd_hexdump, and binwalk_analyze which can be used to read the content of sensitive files if requested.
  • PROMPT_INJECTION (LOW): The skill is susceptible to Indirect Prompt Injection (Category 8) because it processes untrusted binary files and external data with tools that have high-capability access.
  • Ingestion points: strings_extract, ghidra_analysis, binwalk_analyze in references/tools.md.
  • Boundary markers: None identified in tool definitions or workflows.
  • Capability inventory: execute_command, pwntools_exploit, gdb_peda_debug, gdb_analyze in references/tools.md.
  • Sanitization: None identified; tools pass input directly to underlying engines.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 18, 2026, 05:57 PM