The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill utilizes basic Bash commands such as test -f and grep to check for file existence and search for specific text patterns (e.g., 'TODO', 'NEEDS CLARIFICATION'). These operations are constrained to local file inspection within the project's context.
[DATA_EXFILTRATION] (SAFE): No network-capable tools like curl, wget, or fetch are requested or used. The skill remains entirely local to the agent's environment.
[INDIRECT PROMPT INJECTION] (LOW): The skill is designed to ingest and analyze user-provided files (PRDs, SDDs, and source code). While it lacks explicit boundary markers or sanitization logic for these external inputs, its primary mechanism involves searching for static text markers rather than executing instructions found within those files.
[PROMPT_INJECTION] (SAFE): The instructions are purely functional and advisory. There are no attempts to override the agent's safety protocols or system prompts.

specification-validation