Pick Next Shell

Pick Next Shell

Pick the next shell from .turbo/shells/ whose dependencies are satisfied, then carry it through the planning pipeline: expand → refine → self-improve → halt.

Task Tracking

At the start, use update_plan to track each step, restating any remaining steps of a parent workflow alongside them:

• Scan shells and pick next
• Run $expand-shell skill
• Run $refine-plan skill
• Run $self-improve skill
• Mark plan ready
• Summarize and halt

Step 1: Scan Shells and Pick Next

Check terminal conditions first:

• No shells and no plans — nothing to pick; stop
• No shells, but plans exist with status: done for all — the project is complete; stop
• No shells, but plans exist with a non-done status — there are unfinished plans; stop

If shells exist in .turbo/shells/, glob .turbo/shells/*.md and read each file's YAML frontmatter. A shell's depends_on entry is satisfied when .turbo/plans/<dep-slug>.md exists with status: done in its frontmatter.

• Candidates — shells whose depends_on are all satisfied
• Blocked — shells with one or more unsatisfied depends_on

If there are no candidates (everything is blocked), report which shells are blocked and which dependencies they're waiting on, then stop.

If multiple candidates exist, pick the one with the lowest shell number (from the NN- prefix $draft-shells gives each file). If ambiguous, use request_user_input to let the user choose.

State the picked shell path and its dependencies before continuing.

Workflow state lives at .turbo/workflows/<slug>.md — slug from the picked shell's basename without .md. It pairs one-to-one with the thread's goal. When this run's create_goal attempt succeeds, write the file fresh: Status: active plus this invocation's update_plan list as a checkbox list. When an unfinished goal already exists, mirror into the workflow file its objective names; when it names none, continue without workflow state. Mirror every update_plan call into the file; it holds the pipeline's remaining steps and their statuses. When this run created the goal, run the terminal step in order: mark the final entry completed and mirror it, set Status: closed, mark the goal complete with update_goal, then emit any halt message.

Attempt create_goal with the objective: "Carry the shell at <shell path> through expand, refine, and self-improve, mark the plan ready, and halt after the summary. Workflow state: .turbo/workflows/<slug>.md; mirror every update_plan call into it. Loop state lives under .turbo/loops/. After any context compaction, re-read the workflow file and any active ledger, and continue from the first unfinished entry. Mark this goal complete at the halt that follows the plan-ready summary." If an unfinished goal already exists, an outer workflow owns it; continue without creating one.

Step 2: Run $expand-shell Skill

Run the $expand-shell skill, passing the shell file path. Capture the resulting plan path for Step 3.

Step 3: Run $refine-plan Skill

Run the $refine-plan skill with the plan path from Step 2.

Step 4: Run $self-improve Skill

Run the $self-improve skill to compound planning learnings.

Step 5: Mark Plan Ready

Update the plan's YAML frontmatter to status: ready.

Step 6: Summarize and Halt

Present a brief summary of the finished plan: the essence of what it builds and the key decisions behind it, short enough to read at a glance so the user does not have to open the full plan file. When the plan delivers value to a user, developer, or operator, also present a short list of stories capturing what that person gains, in the form "As a <persona>, I want <capability> so that <outcome>". Skip the stories only when no beneficiary or outcome can be named, such as a purely mechanical refactor. Fit both to the plan rather than a fixed template.

If this run created a goal in Step 1, mark it complete with update_goal. Then halt with this message:

> Plan ready at <plan path>.

>

> This is a good point to /compact before implementing. Then run $implement-plan <slug> to implement. After that, run $pick-next-shell again for the next shell.

Rules

• Do not edit plan files directly. Revisions go through $refine-plan.
• Never modify the spec file.
• Do not attempt to auto-implement. The user drives implementation with $implement-plan after an optional /compact.
• If a shell file is missing or has invalid frontmatter, halt and report.