The Agent Skills Directory

[PROMPT_INJECTION]: No malicious instructions or bypass patterns were detected. Instructional shortcuts like /fast and /assume are used purely for task-specific logic and do not attempt to override the underlying agent's safety protocols.- [DATA_EXFILTRATION]: The skill does not perform any network requests. File writing is protected by explicit rules that reject absolute paths, parent directory traversal (..), and sensitive system paths.- [COMMAND_EXECUTION]: A local verification script (scripts/check_sdd_structure.py) is used to validate markdown headings. The script uses only standard Python libraries and does not execute arbitrary shell commands or access untrusted external resources.- [REMOTE_CODE_EXECUTION]: The skill does not download or execute code from remote sources. All logic and references are contained within the local skill repository.- [INDIRECT_PROMPT_INJECTION]: While the skill processes external project requirements and existing design documents, the risk is mitigated by its restricted capabilities. It only writes markdown files to designated project subdirectories and lacks network access for data exfiltration.

software-design-doc