The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute various CLI tools including git (for status, diff, and cloning), just (for running project recipes), copier (for template updates), and rg (for conflict detection). These commands are standard for managing project updates and template synchronization.
[DYNAMIC_EXECUTION]: The skill utilizes the copier tool, which renders Jinja2 templates and can execute shell commands via migrations, tasks, and extensions like jinja2_shell_extension. The instructions explicitly warn the agent to inspect the code and configuration invoked by these features before using the --trust flag, providing a measure of control over dynamic execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from upstream template repositories through Git logs, diffs, and Copier configuration files. A malicious upstream template could attempt to influence the agent's behavior during the update process. The skill provides a mandatory audit workflow for upstream changes to mitigate this surface.
Ingestion points: Upstream template repositories accessed via _src_path, Git logs, and diffs (referenced in SKILL.md and references/python-templates.md).
Boundary markers: The skill mandates auditing the upstream change range and reviewing matches for legitimate content vs markers.
Capability inventory: Shell execution via just, copier, and git commands throughout the update and validation workflow.
Sanitization: Instructions require manual inspection of code, extensions, tasks, and migrations before execution.

copier-template-sync