The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the subprocess module in scripts such as run_eval_claude.py and run_eval_codex.py to invoke local CLI tools with user-defined queries and skill content.
[EXTERNAL_DOWNLOADS]: The skill manages dependencies via uv and fetches the SheetJS library from a public CDN to support spreadsheet rendering in the evaluation viewer.
[PROMPT_INJECTION]: By processing and iterating on user-provided SKILL.md and evals.json files, the skill is exposed to indirect prompt injection where instructions within the skill-under-development could attempt to override the agent's behavior during testing or optimization.

skill-forge