The Agent Skills Directory

[PROMPT_INJECTION]: No override patterns or safety bypass attempts were detected. The instructions strictly enforce boundaries between refactoring and other tasks like features or bug fixes.- [DATA_EXFILTRATION]: No network operations or access to sensitive system files were identified. The skill's operations are confined to the project's source code and a dedicated logging directory.- [REMOTE_CODE_EXECUTION]: The skill does not download or execute external scripts. It relies on local project tools (tests, linters) and manual verification.- [CREDENTIALS_UNSAFE]: No hardcoded secrets or patterns suggesting credential harvesting were found.- [COMMAND_EXECUTION]: Local command execution is limited to standard development tasks (running tests, type checking) and is explicitly documented as part of the refactoring validation workflow.- [INDIRECT_PROMPT_INJECTION]: While the skill processes external data (source code), the risk is mitigated by a multi-stage approval process where a human must review the scan results, the execution design, and each individual code change.

easysdd-refactor