ultracode

ultracode-edit

Safely modify an initialized UltraCode repository configuration and its managed Codex or Claude Code projections. Use when the user explicitly invokes `$ultracode-edit`, wants to change project AI rules, commands, control visibility, status persistence, model or objective-driven reasoning policy, safety cap, roles, skills, or adapters, or needs to repair configuration drift without losing manual edits. Do not select this skill when the request explicitly invokes `$ultracode-help`; `edit` is then only a Help topic.

ultracode-flow

Concise, read-only live control view for an UltraCode task. Use when the user explicitly invokes `$ultracode-flow` or asks for a quick but fully explained view of the objective, phase, active tickets, assigned agents, requested and effective models and reasoning effort, blockers, completion criteria, and next actions. Do not select this skill when the request explicitly invokes `$ultracode-help`; `flow` is then only a Help topic. Use `full` for all known tickets or `agents` for an agent-focused view.

ultracode-help

Explain and help choose UltraCode commands without changing the task or repository. Use when the user invokes `$ultracode-help` or `ultracode-help`, asks how UltraCode works or which command fits, or asks about UltraCode initialization, feature verification, control, tickets, agents, models, reasoning effort, safety, validation, or examples. Support focused command and model topics plus an explicitly brief or synthetic mode.

ultracode-init

Guided, safe initialization of UltraCode project control for Codex and Claude Code. Use when the user explicitly invokes `$ultracode-init`, asks to adopt UltraCode in a repository, or wants shared AI documentation, rules, skills, reviewers, model and objective-driven reasoning policy, configuration, and status structure generated from the real project without copying machine-local settings. Do not select this skill when the request explicitly invokes `$ultracode-help`; `init` is then only a Help topic.

ultracode-status

Detailed, read-only diagnosis of an UltraCode task. Use when the user explicitly invokes `$ultracode-status` or asks for complete job history, files, findings, model and reasoning effort, validation evidence, blockers, persisted-state freshness, or a full diagnostic account. Do not select this skill when the request explicitly invokes `$ultracode-help`; `status` is then only a Help topic. Prefer `$ultracode-flow` for a quick current-work snapshot. Never mutate the repository merely to refresh status.

ultracode-verify

Create, inspect, execute, and maintain a durable feature-level functional verification plan with evidence-backed scenario results. Use when the user explicitly invokes `$ultracode-verify`, asks to prove that a feature works, wants reusable acceptance scenarios, or needs a persistent verification record. Use only the statuses planned, passed, failed, not-run, and not-applicable; fail closed on incomplete or inconsistent evidence. Do not select this skill when the request explicitly invokes `$ultracode-help`; `verify` is then only a Help topic.

Code & Development Listed

ultracode

Adaptive, user-visible software-engineering orchestration for Codex. Use when the user explicitly invokes UltraCode or requests a complex implementation, difficult diagnosis with a requested fix, broad audit, refactor, migration, reverse-engineering effort, or other work that benefits from evidence-driven decomposition, many bounded agent jobs, adversarial verification, real validation, and one coherent synthesis. Do not select this skill when the request explicitly invokes `$ultracode-help`; any following command name is a Help topic. Do not invoke implicitly for simple questions, trivial localized edits, or work fully owned by a narrower specialist skill.