zgeoff

Boot a vers domain service as a real process and drive its RPC endpoints over HTTP — for changes the in-process test client bypasses (serve entrypoint, env wiring, transport middleware, error serialization, trace propagation).

Writing rules for repo prose — AGENTS.md and agents/ partials, docs/, README content, and doc comments. Load before writing or editing any of them.

Write specifications for AI agents following a specialized template that separates contract-level content from prompt-level content. Use when creating agent specs, specifying subagent behavior, or documenting agent contracts. Triggers on requests like "write an agent spec", "spec out this agent", "create a spec for the X agent", or "document the agent's behavior". Invokes /doc-coauthoring for structured collaboration.

Write project specifications following standard templates, writing disciplines, and conventions. Use when creating specs, technical specifications, feature specs, or system design documents. Triggers on requests like "write a spec for X", "create a specification", "spec out this feature", or "document the requirements for Y". Invokes /doc-coauthoring for structured collaboration.

Write specifications for AI agents following a specialized template that separates contract-level content from prompt-level content. Use when creating agent specs, specifying subagent behavior, or documenting agent contracts. Triggers on requests like "write an agent spec", "spec out this agent", "create a spec for the X agent", or "document the agent's behavior". Invokes /doc-coauthoring for structured collaboration.

Internal agent skill for GitHub Issue and PR operations via the `gh` CLI. Used by workflow agents (Planner, Implementor, Reviewer) for: issue CRUD, PR lifecycle (create, review, merge), label management with mutually exclusive categories and status transitions, comment posting, CI status checks, and query patterns. Not user-invoked; triggered internally when agents perform GitHub operations defined in the development protocol.

Rewrites words and phrases commonly produced by LLMs into plainer, more common alternatives within one or more files. Use when the user asks to "declaudeify", "de-LLM", de-slop, or plain-language a file, or to reword telltale AI phrasing. If no file is named, it asks which file to operate on.