The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were detected in the skill instructions or templates.
[COMMAND_EXECUTION]: The skill performs local command execution to run the tests it generates. This behavior is standard for a testing tool and is confined to the project's development environment.
[REMOTE_CODE_EXECUTION]: While the skill involves execution of generated code, it does not involve fetching or executing code from remote or untrusted internet sources. All execution is local and based on project code analysis.
[PROMPT_INJECTION]: The skill processes implementation diffs as input. The risk of indirect prompt injection is mitigated by the skill's focus on technical analysis and code generation rather than general instruction following from the input data.

adversarial-tester