The Agent Skills Directory

[PROMPT_INJECTION]: A regex hit was identified on instructions starting with 'Untrusted-content envelope...'. This is a false positive for malicious injection; it is a defensive instruction advising the model to treat data from instrumented targets strictly as data and ignore any embedded commands within that untrusted content.
[REMOTE_CODE_EXECUTION]: The skill documents the installation of frida-tools via pipx. This is a standard installation of a well-known security tool and is considered safe practice for its intended purpose.
[COMMAND_EXECUTION]: The skill facilitates the execution of Frida commands and custom instrumentation scripts. These are necessary for the primary functionality of the skill (dynamic analysis) and include clear guidance on usage and sandboxing profiles.

frida