The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute local repository scripts for tasks such as generating indices and validating file references. These operations are limited to project-specific directories (agents/, scripts/) and represent routine automated maintenance tasks.- [DYNAMIC_EXECUTION]: Inline Python snippets are used via python3 -c to perform YAML validation and file existence checks. These operations use standard libraries and safe parsing methods to ensure that newly created agent files adhere to the required schema.- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from user-provided keywords and existing repository files (ingestion points). It mitigates injection risks by defining a strict instruction hierarchy in its design patterns, which explicitly treats retrieved content as data rather than instructions (boundary markers). While the skill possesses file-write and execution capabilities (capability inventory), it enforces mandatory structural and framing validation through automated scripts (sanitization).

agent-creator