exploitability-validation
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to run various internal command-line tools located in
libexec/(e.g.,raptor-validation-helper,raptor-run-sandboxed,raptor-smt-validate-path). These tools are used for pipeline orchestration, inventory building, and binary analysis. - [DYNAMIC_EXECUTION]: The pipeline involves compiling and executing Proof of Concept (PoC) code to verify vulnerabilities. The instructions strictly mandate the use of a sandbox (
libexec/raptor-run-sandboxed) for all compilation and execution tasks to restrict network access and file system writes. - [INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to process untrusted source code and vulnerability findings. It implements a defensive 'Untrusted-content envelope' strategy, explicitly instructing the agent to treat target content as data and disregard any embedded commands such as 'ignore previous instructions' or 'run this command'.
- [PROMPT_INJECTION]: While automated scanners flagged patterns like 'ignore previous instructions', these occurrences are part of defensive guidelines that warn the agent against following instructions found within analyzed target code, rather than attempts to override the agent's core safety protocols.
Audit Metadata