The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of prompt injection or instructions to bypass safety guidelines was found. The skill provides structured guidance for code auditing without attempting to override the agent's core instructions.
[DATA_EXFILTRATION]: The skill does not contain any network operations, hardcoded credentials, or access to sensitive system file paths. It explicitly states that analysis is performed offline on provided source code.
[REMOTE_CODE_EXECUTION]: There are no patterns involving the download or execution of remote scripts. The skill relies entirely on the model's native ability to analyze text.
[COMMAND_EXECUTION]: No shell commands or subprocess execution patterns were identified in the instructions or reference files.
[OBFUSCATION]: The content is provided in clear text with no evidence of Base64 encoding, zero-width characters, or other obfuscation techniques.
[INDIRECT_PROMPT_INJECTION]: While the skill is designed to process untrusted data (source code for auditing), it defines a strict multi-pass procedure to ensure the model focuses on authorization logic. It does not have access to sensitive capabilities like network or system-level write access, limiting the potential impact of adversarial content within the audited code.
[DYNAMIC_EXECUTION]: The skill contains no dynamic code generation, runtime compilation, or unsafe deserialization logic.

authz-security