agentleFS
Data-WiseData-Wise/craft/skills/orchestration/drive-engine/SKILL.md

drive-engine

Skill4 starsChanged 12 days agoMITView on GitHub

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.

  • Commits and pushes

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Keep the transcript linear when N == 1 so the /goal evaluator can read it.

    line 19
  2. Very specific:

    Look for, in order: worktree ORCHESTRATE-*.md → SPEC ## Phase N / ## Increment N headings → top-level numbered task list.

    line 28

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Drive Engine

The reusable execution body behind /craft:orch:drive. The command owns condition synthesis and gating; this skill owns the work.

Responsibilities

  1. Resolve phases — if an 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.
  2. Dispatch — per turn, launch --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.
  3. Real verify gate — when the /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.

Phase resolution

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[]}.

Verify-command auto-detection

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.

Outputs

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.

Not this skill's job

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.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.