The Agent Skills Directory

[Command Execution]: The skill uses shell commands for environment discovery, such as checking git config user.name and the GIT_AUTHOR_NAME variable. These are standard developer workflow operations used specifically for document attribution and do not present a security concern in this context.
[Interaction Invariants and Constraints]: The skill implements strict interaction rules, such as mandatory artifact-first delivery. This ensures large blocks of text are written to files rather than filling the chat interface, which improves usability and maintains workspace organization.
[Trust and Consensus]: The workflow includes mandatory 'Approval Gates' at each step, ensuring the user reviews and confirms all technical decisions and draft sections before the agent proceeds. This provides high transparency and user control.
[Data Ingestion and Reconciliation]: The skill includes a reconciliation step (Step 6) that compares the final draft against the initial concepts brief to detect contradictions. This is a safety feature that prevents the agent from deviating from established requirements without user notification.

egm-draft