annotations
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
Bashtool to executeradare2commands for analyzing and mutating binary data. This grants the agent the capability to interact with the host filesystem and binary analysis environment via the command line. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill ingests untrusted data derived from binary analysis, such as function names, strings, and existing comments, which are processed during the
r2xsqlmutation workflows inSKILL.md. - Boundary markers: The skill documentation specifies that identifier names are restricted to the
[A-Za-z0-9._$]character set and variable types are validated against a specific regex to mitigate injection risks. - Capability inventory: The skill can execute shell commands through the
Bashtool and perform filesystem writes via radare2's project saving mechanism (Ps). - Sanitization: The documentation describes the use of a structural command-dispatch path that avoids the evaluation of radare2 command separators, acting as an internal layer of defense against malicious input during annotation updates.
Audit Metadata