The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local system commands and scripts for validation and asset processing. This includes running a Python validation script located at python3 "${CODEX_HOME:-$HOME/.codex}/skills/.system/skill-creator/scripts/quick_validate.py", as well as standard tools like npm test, git diff, xmllint, and cwebp.
[DATA_INGESTION]: The skill processes user-provided ideas and workflows to generate new SKILL.md files and automation scripts. This creates a surface for indirect prompt injection, where a malicious user prompt could influence the logic of the generated skill. However, the skill instructions emphasize following established repository conventions and performing multi-step validation.
Ingestion points: User instructions and workflow ideas provided at runtime.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are mandated for the generated content.
Capability inventory: The skill is capable of writing files and executing local subprocesses including Python, Node.js/NPM, Git, and image processing tools.
Sanitization: No specific sanitization or escaping logic is defined for the interpolated user content.
[DYNAMIC_EXECUTION]: The skill's primary function is the generation of new executable content, including Markdown-based agent instructions and automation scripts within the scripts/ directory of the newly created skill. This is the intended behavior for a scaffolding tool.

create-skill