The Agent Skills Directory

[SAFE]: The skill focuses exclusively on the reduction and deletion of redundant code, minimizing the risk of introducing malicious logic or backdoors.
[PROMPT_INJECTION]: The skill analyzes external content (the codebase being audited), which could contain adversarial instructions. However, the skill is protected by robust guardrails: it forbids code additions, style rewrites, and modifications to lockfiles or vendored code. Furthermore, every proposed deletion must be reviewed and approved by the user before execution, which prevents the agent from following instructions hidden within the source files.
[COMMAND_EXECUTION]: The skill instructs the agent to run existing project tests or build commands to verify that deletions have not impacted system stability. This is a standard development practice and relies on the user's local configuration.

fstack-simplify