The Agent Skills Directory

[COMMAND_EXECUTION]: The Python script scripts/repair_codex_skill_format.py is used to discover and update skill configurations. It executes git rev-parse to identify project roots. The skill also references a hardcoded validation script path /Users/khang/.codex/skills/.system/skill-creator/scripts/quick_validate.py.
[PROMPT_INJECTION]: The skill processes content from various skill directories, creating a surface for indirect prompt injection.
Ingestion points: The repair script reads SKILL.md files from user-scope and project-scope directories including ~/.codex/skills and ~/.claude/skills.
Boundary markers: No explicit delimiters or instructions are used to distinguish between skill metadata and potentially malicious instructions within those files.
Capability inventory: The skill can create directories, write files, and execute shell commands via Python's subprocess and Path modules.
Sanitization: The script uses yaml.safe_load for parsing YAML frontmatter and applies regex-based normalization to names and descriptions.

fix-codex-skill