The Agent Skills Directory

[SAFE]: The skill's operations, including file creation via the Write tool and the execution of local validation scripts (scripts/build-codex-agents.py, scripts/validate-agents.py, scripts/validate-marketplace.py), are strictly aligned with its documented purpose as an agent development utility. No unauthorized network activity or credential exposure was found.
[COMMAND_EXECUTION]: The skill automates the registration of new agents by running local Python scripts to build Codex definitions and validate marketplace metadata. These commands are executed within the local repository context and are part of the intended workflow.
[INDIRECT_PROMPT_INJECTION]: The skill takes user input to populate agent system prompts. It mitigates potential injection risks by providing structural templates (Input, Process, Rules, Output) and requiring manual confirmation at each of the six development phases.

agent-creator