The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches structural improvements, manifests, and skill files from the vendor's remote repository at https://github.com/Rheinmir/setup.git and https://raw.githubusercontent.com/Rheinmir/setup/master/.
[COMMAND_EXECUTION]: The skill invokes local Python scripts (harness/scripts/sync-template.py, harness/scripts/health-check.py, harness/scripts/okf-check.py) and uses system binaries like curl and python3 to perform synchronization and health checks.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from a remote repository and interpolates it into the local environment.
Ingestion points: Remote files fetched from the Rheinmir/setup GitHub repository.
Boundary markers: None explicitly implemented to delimit remote content during processing.
Capability inventory: File system write access to project directories and user-level agent configuration paths; network access via curl and gh api.
Sanitization: The skill contains rules (RULE-01) to explicitly skip sensitive files such as .env and credentials.
[DYNAMIC_EXECUTION]: The skill generates and modifies skill and command files by prepending YAML frontmatter to downloaded content before installing them into agent-specific configuration paths (.claude/commands/, ~/.claude/skills/).
[PERSISTENCE]: The skill installs synchronized files into the agent's global configuration directories (~/.claude/skills/ and ~/.claude/commands/), ensuring the downloaded logic persists across different agent sessions and projects.

sync-template