The Agent Skills Directory

[SAFE]: The skill is purely instructional and contains no scripts, binaries, or executable code that could perform unauthorized actions on the host system.
[NO_CODE]: No Python or Node.js dependencies are defined, and no shell commands are present in any of the skill files.
[SAFE]: The instructions include a specific safeguard against indirect prompt injection by telling the agent to treat the translation as a reading aid and to ignore any questions or proposed tasks within the content being translated.
[SAFE]: The configuration in the YAML frontmatter and the openai.yaml file explicitly disables autonomous invocation, meaning the skill can only be triggered by a direct user request.

traduzir