The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard shell utilities like ls and grep to verify the existence of files and the presence of required structural markers. These commands are performed on local paths provided by the user and do not involve shell piping to interpreters or external network requests.
[PROMPT_INJECTION]: No malicious instruction overrides or safety bypass attempts were detected. The use of terms like "IMPORTANT RULES" and "EXECUTION INSTRUCTIONS" refers to structural headers within the prompt files being validated for formatting compliance, rather than instructions to the agent to ignore its core safety guidelines.
[DATA_EXFILTRATION]: No network-facing code or patterns suggesting the collection of sensitive environment data or credentials were found. All file operations, including reading task files and writing backups during the repair process, are confined to the local directory structure defined by the user.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute remote scripts. The --fix functionality regenerates local files using a predefined template and existing local task descriptions, posing no risk of remote code injection.

parallel-validate-prompts