================ Workflow Skills ================ When you run ``moot init``, it quietly installs seven files into ``.claude/skills/``. This page explains what they are and why they are there. What a skill is --------------- A skill is a set of instructions an agent follows when working on a specific task. Think of it as a playbook: structured guidance that tells an agent how to approach a particular kind of work — how to write a spec, how to pass work to the next agent, how to verify a feature. Skills live in ``.claude/skills/`` in your project directory. Each skill is a folder containing a ``SKILL.md`` file. Agents load the relevant skill at startup or when they need it. You do not invoke skills yourself — the agents pick them up automatically, which is what makes agent behavior consistent across sessions and across team members. The seven bundled skills ------------------------ ``product-workflow`` ~~~~~~~~~~~~~~~~~~~~ Guides the Product agent through feature planning, kickoff composition, and retro synthesis. When the Product agent starts a new feature, it follows this skill to scope the work, compose the kickoff message that hands off to the rest of the team, and synthesize retrospectives into lasting team memory. ``spec-checklist`` ~~~~~~~~~~~~~~~~~~ Helps the Spec agent write complete, consistent design specs. It walks through a structured checklist covering API changes, data model impacts, frontend effects, test requirements, and security considerations. A spec produced with this checklist gives the rest of the team a reliable blueprint to build from. ``leader-workflow`` ~~~~~~~~~~~~~~~~~~~ Guides pipeline orchestration. The Leader agent uses this skill to manage handoffs between agents: creating feature branches, monitoring progress, merging branches at the right time, and shipping completed work. Without this skill, handoffs would depend on the agent remembering the right sequence of steps. ``librarian-workflow`` ~~~~~~~~~~~~~~~~~~~~~~ Guides documentation maintenance and as-built passes. The Librarian agent uses this skill to update ``docs/design/``, ``docs/arch/``, and related docs after a feature ships — making sure the documentation reflects what was actually built, not just what was planned. ``handoff`` ~~~~~~~~~~~ Structures how agents pass work to each other. Every agent uses this skill when completing their phase of a feature: what to commit, what to say in the handoff message, who to notify, and when to stop and wait. Consistent handoffs are what keep the pipeline moving without agents polling or second-guessing each other. ``verify`` ~~~~~~~~~~ Guides the QA agent's verification workflow. When QA receives a handoff, this skill walks through rebuilding the test stack, running backend tests, checking the frontend build, reviewing the implementation against the spec, and posting a structured pass/fail report. It also covers when QA can make small repairs directly versus when to escalate. ``doc-curation`` ~~~~~~~~~~~~~~~~ Guides documentation restructuring and maintenance. Used when docs need reorganization: deciding when to split or merge files, adding cross-references, reducing duplicated explanations, and keeping directory indexes current. This skill helps maintain a doc tree that agents can navigate reliably. How skills work --------------- Agents load skills at startup or on demand. The Skill tool in Claude Code lets an agent invoke a skill by name — for example, the Leader agent calls ``leader-workflow`` at the start of every session. You do not need to do anything to activate skills. Once they are in ``.claude/skills/``, agents discover and use them automatically. This is what makes agent behavior predictable: the same skill file governs the same behavior whether the session started today or three months ago. Customizing skills ------------------ Skills are plain text files. You can open any skill in ``.claude/skills/`` and edit it to match your team's conventions. Changes take effect the next time an agent starts a session — no restart or reinstall required. You can also create new skills for tasks specific to your project. Add a folder with a ``SKILL.md`` file and agents will be able to load it by name. If ``moot init`` detects existing ``.claude/skills/`` content, it writes suggested updates to ``.moot/suggested-skills/`` and creates an init report rather than overwriting your files. This lets you review and merge changes on your own terms. Harness compatibility --------------------- Skills as slash commands are a Claude Code feature. If you are using a different editor — Cursor, Cline, or another MCP-compatible harness — the skill files are still installed in ``.claude/skills/`` and agents can read them directly. The agent behavior guided by skill content works regardless of harness. What may not be available is invoking a skill as a slash command (``/leader-workflow``) from the editor's command palette; check your editor's documentation for how it handles custom slash commands.