The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/scaffold.py script executes system commands such as gh repo create, gh repo edit, git init, and git remote to manage the lifecycle of the scaffolded repository.
[COMMAND_EXECUTION]: The provided template includes safety and validation hooks located in .claude/hooks/ (guard_bash.py and validate_skill_file.py). these scripts are executed by the agent platform during tool use (Edit/Write/Bash) to enforce repository policies.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied input for the repository name, description, and idea. This data is interpolated into various files, including the .local/PROMPT.md file. Because this file is intended to be used as a goal prompt for the agent in the new repository, malicious input could influence the behavior of the agent in that new context.
Ingestion points: Arguments name, description, and idea in SKILL.md.
Boundary markers: None present; the user idea is placed directly into markdown templates.
Capability inventory: The generated environment includes subprocess execution (hooks) and network access (deployment skills).
Sanitization: scaffold.py uses json.dumps for JSON manifest files, but uses raw string replacement for markdown and documentation files.
[EXTERNAL_DOWNLOADS]: The publish-repo skill in the template utilizes npx skills@latest to install and verify the repository's skills, which involves downloading packages from the official NPM registry.
[DYNAMIC_EXECUTION]: During the scaffolding process, scaffold.py executes the newly generated scripts/validate_skills.py script to verify the integrity and compliance of the fresh scaffold before completing the operation.

create-skill-repo