The Agent Skills Directory

[COMMAND_EXECUTION]: Several scripts in the scripts/ directory utilize the subprocess.run() module to perform essential development tasks. Examples include sync_mise_primitives.py (executing mise --version), check_code_rules.py (invoking linters), and agentic_request_contract.py (running unit tests). These are standard operations for a development-oriented tool.
[EXTERNAL_DOWNLOADS]: The script scripts/sync_mise_primitives.py fetches official JSON schemas from the jdx/mise repository on GitHub (raw.githubusercontent.com). This download is performed to keep the skill's local primitive catalog synchronized with the external mise tool version. The target domain is a well-known service and the content is limited to configuration schemas.
[DYNAMIC_EXECUTION]: The skill's apply command, implemented in scripts/lib/apply_plan.py, allows for the execution of validation commands defined within an integration plan. This mechanism is used to verify that modifications to an agent skill (such as inserting new behavior rules) do not break the target skill's existing test suite or static checks. The execution is confined to the target skill's directory.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface through user-supplied behavior configurations (in Markdown, YAML, or JSON) which are then interpolated into generated SKILL.md files. While this creates a potential surface for instructions to be passed through, the skill mitigates risk by requiring domain-specific terminology through scripts/domain_text.py and strictly using templates for code generation, ensuring the resulting artifacts follow a predictable and safe structure.

tool-call-configuration-for