The Agent Skills Directory

[SAFE]: The skill's primary focus is on 'adversarial verification,' instructing the agent to actively refute its own findings and use fresh context passes to ensure implementation quality. This defensive reasoning approach reduces the risks associated with processing potentially untrusted codebase data.
[COMMAND_EXECUTION]: The workflow includes running project tests to verify the grounded outcome of implementation tasks, which is an expected and legitimate function for a software engineering skill.
[DATA_EXFILTRATION]: Analysis of file system operations shows that the skill reads project context and saves its results to a standard local directory (~/.agents/output/). There are no unauthorized network operations or attempts to access sensitive system credentials.

ultrapex