菜单
This skill should be used when driving an approved SPEC to completion — "drive this spec to done", "run the orchestrate drive loop", "implement the spec autonomously until tests pass". Owns the reusable body behind /craft:orch:drive — parse-or-derive ORCHESTRATE phases, dispatch file-scoped subagents, and run the authoritative real verify gate.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
This skill should be used when the user asks to "pre-flight check", "validate before commit", "validate before PR", "validate before deploy", "is this ready to ship", "check readiness", "run pre-flight", "validate project state", or wants context-aware validation across commit/PR/release/deploy boundaries. Orchestrates universal validators (lint, tests, types, version sync, stale refs, docs, hook conflicts) and supports generating new custom validators.
This skill should be used when a command's --refine flag is set, or the user asks to "refine my prompt", "optimize this prompt", "make this request sharper" — rewrites a vague natural-language request into a specific, well-structured prompt using project context, shows before/after, and confirms before the caller proceeds. Replaces the deprecated /craft:refine command.
This skill should be used when the user asks to "release", "create a release", "ship it", "publish a new version", "bump version and release", "cut a release", or mentions releasing to GitHub/npm/PyPI. Orchestrates the full release pipeline from pre-flight checks through GitHub release creation.
ADHD-friendly session and task workflow for craft users — captures context at session end, restores it at session start, picks the next task, breaks through blockers, holds focus on one task, reviews a craft spec, or refines a vague craft prompt. Use when wrapping up or resuming a coding session, when deciding what to work on next (also "what's next", "pick a task for me", "give me something to do"), when stuck or in decision paralysis, when focusing on a single craft task, or when reviewing/refining a craft spec or prompt. The full intent-to-operation map is in the "When to Use" table below.
This skill should be used when the user asks to "init a repo", "set up git", "manage branches", "switch branch", "create a worktree", "list worktrees", "clean merged branches", "sync with remote", "git activity", "git recap", "protect main", "apply branch protection", "bypass protection", "undo a commit", "fix a git mistake", or mentions git lifecycle, worktrees, branch protection, GitHub-side baseline protection, safety rails, or learning git workflows. Covers craft's full git lifecycle — repo init, branch and worktree management, remote sync, status/recap, local and GitHub-side branch protection, and the learning/safety reference docs.
This skill should be used when the user asks "what's my friction", "session insights", "friction report", "what patterns am I hitting", "analyze my sessions", "generate insights report", "aggregate facets", or "show CLAUDE.md suggestions". Aggregates session facet history into a friction/goals report. Formerly bundled with the brainstorm skill (ideation); split out because the two operations share no input or output. Directory name kept as brainstorm-insights for path stability — tests and references/insights.md reference it.
| name | drive-engine |
| description | This skill should be used when driving an approved SPEC to completion — "drive this spec to done", "run the orchestrate drive loop", "implement the spec autonomously until tests pass". Owns the reusable body behind /craft:orch:drive — parse-or-derive ORCHESTRATE phases, dispatch file-scoped subagents, and run the authoritative real verify gate. |
The reusable execution body behind /craft:orch:drive. The command
owns condition synthesis and gating; this skill owns the work.
ORCHESTRATE-*.md exists in the worktree
root, parse its phases and file-scopes. If not, derive phases directly
from the SPEC's ## Phase/## Increment/task-list sections. An
ORCHESTRATE file is preferred when present but never required.--agents N (default 1) file-scoped
subagents, one per pending wave item, each scoped to its files only.
Keep the transcript linear when N == 1 so the /goal evaluator can
read it./goal condition clears, run the
project's actual verification (auto-detected, see table) and treat its
exit status as the authoritative "done". A green transcript is NOT
sufficient; the command must actually run.Look for, in order: worktree ORCHESTRATE-*.md → SPEC ## Phase N /
## Increment N headings → top-level numbered task list. Emit a wave list
of {phase, files[], tasks[]}.
See ../references/verify-gate-detection.md for the
full detection table and how to use it. Always pair with git status --short to confirm a clean,
committed tree.
A structured verify result: { command, exit_code, passed: bool, summary }.
On passed: true, the caller stops at verified-green and prints the
gh pr create command — this skill never opens a PR.
The #NNN issue-premise pre-filter (does the spec cite an open issue? if
so, run /craft:git:issue-check <N> and surface the verdict) lives in
/craft:orch:drive's own Step 2, not here — it's part of the command's
gating responsibility, run once before any dispatch begins. It never
blocks and never runs unless a citation is found.