skills/zeij-drive/deobf/deobf-all/Gen Agent Trust Hub

deobf-all

Pass

Audited by Gen Agent Trust Hub on Jul 1, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill functions as a coordinator, providing instructions to load other skills and follow specific checklists for reverse engineering. It does not contain any executable scripts or direct system commands.
  • [PROMPT_INJECTION]: The skill is designed to process untrusted data (obfuscated code and binaries) provided by users, which represents a surface for indirect prompt injection.
  • Ingestion points: Processes external code targets such as native binaries, JavaScript, and bytecode (SKILL.md).
  • Boundary markers: There are no explicit delimiters or 'ignore-previous-instructions' wrappers defined to isolate the content of the code being analyzed from the agent's instructions.
  • Capability inventory: The skill orchestrates sub-skills capable of deep analysis, symbolic execution, and binary modification across multiple files.
  • Sanitization: No automated sanitization or filtering of the input code is described in the triage or deobfuscation workflows.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 1, 2026, 06:11 AM
Security Audit — agent-trust-hub — deobf-all