spek-plan

name	spek-plan
description	Create a new Plan from an approved Specification.

STOP. Read this before running any command below. A single successful CLI call — including the very first plan new — is NOT task completion. It is not a milestone to report back to the user. It is one step out of many in a workflow that you must keep driving, turn after turn, without stopping, until the CLI itself tells you the workflow is finished. If you find yourself about to say "successfully completed" or summarize results after calling plan new or plan goto even once, you are wrong — go back and read the instruction field you just received, do what it says, and call goto again.

What this skill does

This skill drives a multi-step interactive workflow that produces a complete implementation plan in .spektacular/plans/<name>.md from an existing spec. The workflow is owned by the {{command}} CLI, not by you — the CLI is the state machine and you are the executor.

On each turn, the CLI returns JSON containing an instruction field. That instruction describes exactly one step (e.g. discovery, data structures, phases, testing approach, …). You must:

1. Read the instruction carefully.
2. Perform the step — this may mean researching the codebase, spawning subagents, interviewing the user, or committing a plan document to the store.
3. When the step is complete, run the goto command named at the bottom of the instruction to advance the state machine.
4. Read the next instruction from the new JSON response and repeat.

This is a loop. Do not stop after the first step. Keep looping — step → goto → next instruction → step — until a returned instruction tells you the workflow is finished. Only then should you report completion to the user.

Concretely: do not stop after plan new. That command only starts the workflow — it returns the first instruction (the overview step), not a finished plan. Seeing a clean JSON response with no error is not a signal to stop; it is the signal to keep going. Reporting success, summarizing "plan initialized," or handing control back to the user at this point is the single most common way this skill is executed incorrectly — do not do it.

Reading and writing plan files

The CLI owns the plan documents — plan.md, context.md, and research.md. Never read or write them with the Write, Edit, or Read tools — those bypass Spektacular and the configured plan directory. All plan document access goes through {{command}} plan file:

• {{command}} plan file read <name>/<doc>.md — read a plan document from the plan store.
• {{command}} plan file write <name>/<doc>.md --from <source-path> — write a plan document into the plan store from a source file on disk. Stage the body under .spektacular/tmp/ first, then rm the scratch file after a successful write.
• {{command}} plan file list — list plans in the plan store.

Path arguments are plan-directory-relative document paths (e.g. my-feature/plan.md); plan file resolves them against the configured plan directory itself.

Working files vs. the store documents

While you gather each section, write that section's agreed content directly to its own git-tracked working file under .spektacular/work/<plan_name>/<section>.md using your own Write tool (the phases step writes two: phases_plan.md and phases_context.md). These working files are not store documents — writing them directly with Write is correct and expected, and is the one deliberate exception to the "never use Write/Edit" rule above. That rule protects only the final assembled plan.md, context.md, and research.md, which are written solely through {{command}} plan file write. The per-section working files are scratch-but-durable: the assemble step reads them back to build the three documents (staged to .spektacular/tmp/), the verification step checks the staged documents, the write steps commit them, and then the working directory is removed once all three store writes succeed.

The working sidecar .spektacular/context.md (at the repo's .spektacular/ root — not the plan's own context.md document) has a narrower role: it holds only your cross-cutting learnings and the answers the user gave to your questions — never a copy of section content (that lives in the per-section working files). On resume, read back both the section working files in .spektacular/work/<plan_name>/ and .spektacular/context.md, so you continue from the interrupted step without re-asking for sections already completed.

How to start

Ask the user which spec to plan against before proceeding. You don't need to look for an in-progress workflow yourself — the CLI detects and reports one for you (see below).

Start the plan workflow by running:

{{command}} plan new --data '{"name": "<spec_name>"}'

If a workflow was interrupted and is still in progress, this command does not start a fresh one. Instead it returns a resume report — a JSON object with "resumable": true plus the in-progress workflow's kind, name, and current_step, and an instruction field — and changes nothing on disk. When you get a resume report:

First check the report's kind. If it is not plan, a different workflow (a spec or implement run) is in progress — you cannot resume it from the plan skill, and the CLI will refuse to. Do not run a plan goto. Instead follow the report's instruction: tell the user a <kind> workflow is in progress and let them choose — continue it with that workflow's skill ({{command}} <kind> goto), or discard it and start the plan with {{command}} plan new --force. Only proceed with the steps below when the report's kind is plan.

1. Ask the user whether to resume the in-progress plan or start a new one. (The report's instruction field restates both options.)

2. To resume, first read back the previous session's work with your own file tools: the per-section working files under .spektacular/work/<name>/ (sections already completed) and .spektacular/context.md (learnings + the user's answers). Then run the resume command using the report's current_step:

   {{command}} plan goto --data '{"step":"<current_step>"}'
   

3. To start fresh (discarding the in-progress workflow — it remains recoverable via git), re-run with --force:

   {{command}} plan new --force --data '{"name": "<spec_name>"}'
   

Otherwise the command returns the first instruction and a fresh workflow has started. From that point on, follow the loop above: do what the instruction says, then call {{command}} plan goto --data '{"step":"<next_step>"}' to get the next one. Do not invent step names — every instruction tells you the exact goto command to run next.