spec
/spec
Author a rulebook task spec interactively — research, draft, ask the user clarifying questions, confirm, then create the tasks in rulebook ready for /rulebook-driver. Use when the user wants to plan/spec a feature before implementing.
- Author
- hivellm ↗
- Repository
- rulebook ↗
- Licence
- Apache-2.0
- Last repository change
- 14.10.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 14.10.2026, without references to retired models or APIs.
Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.
What it does and when to use it
Author a rulebook task spec interactively — research, draft, ask the user clarifying questions, confirm, then create the tasks in rulebook ready for /rulebook-driver. Use when the user wants to plan/spec a feature before implementing.
Original author description
Author a rulebook task spec interactively — research, draft, ask the user clarifying questions, confirm, then create the tasks in rulebook ready for /rulebook-driver. Use when the user wants to plan/spec a feature before implementing.
How it works
- Steps
- Rules
SKILL.md preview
---
name: spec
description: Author a rulebook task spec interactively — research, draft, ask the user clarifying questions, confirm, then create the tasks in rulebook ready for /rulebook-driver. Use when the user wants to plan/spec a feature before implementing.
model: opus
context: main
---
Author a rulebook task spec for: $ARGUMENTS
This skill runs in the MAIN conversation (not a forked subagent) so it CAN ask
the user questions with the AskUserQuestion tool. It drives the `spec-author`
workflow as its drafting/critique engine, loops with the user until the spec is
solid, confirms, then creates the rulebook tasks and hands off to
`/rulebook-driver`.
## Steps
1. **Pre-flight.** If `$ARGUMENTS` is empty, ask the user what they want to spec.
Read `.rulebook/specs/rulebook.md` for the required spec format
(## ADDED/MODIFIED, "### Requirement: <name>" with SHALL/MUST, "#### Scenario:"
with Given/When/Then).
2. **Draft + critique loop.** Run the `spec-author` workflow with the topic:
`Workflow({ name: "spec-author", args: { topic: "<topic>" } })`.
It returns `{ draft, ready, questions[], gaps[], missingScenarios[] }`.
3. **Ask the user.** If `ready` is false, take the returned `questions` (each has
`question`, `why`, `options`) and present them with the **AskUserQuestion**
tool — one question per item, using the provided `options` as choices (the
user can always answer free-form). Also surface `gaps` and `missingScenarios`
so the user can react to them.
4. **Iterate.** Re-run the workflow feeding the answers back:
`Workflow({ name: "spec-author", args: { topic: "<topic>", answers: [{ question, answer }, ...] } })`.
Accumulate ALL answers across rounds (don't drop earlier ones). Repeat steps
3–4 until the workflow returns `ready: true` (no open questions).
5. **Confirm with the user.** Show the final proposal + spec (the workflow's
`draft`). Ask the user to confirm with AskUserQuestion: **"Create these
rulebook tasks?"** options: Create / Revise (more questions) / Cancel.
- Revise → go back to step 3.
- Cancel → stop, leave nothing created.
- Create → continue.
6. **Create the tasks.** ONLY after explicit confirmation, create the task(s) in
rulebook using the MCP tools — never `mkdir`/`Write` by hand:
- `rulebook_task {action:"create"}` for each task (phase-prefixed id, e.g.
`phase1_<slug>`), passing the confirmed `why`, `what`, `items` checklist
and `impact` so proposal.md and tasks.md are written complete; then write
specs/<module>/spec.md from the draft.
- `rulebook_task` each created task; fix format issues and re-validate
until clean.
7. **Hand off to the driver.** Report the created task ids and tell the user the
spec is ready. Offer to start implementation now by running the
`rulebook-driver` workflow: `/rulebook-driver` (drains the whole backlog) or
`/rulebook-driver { "once": true }` for one item. Do NOT auto-start it without
the user's go-ahead.
## Rules
- The user MUST confirm (step 5) before any task is created. No silent creation.
- Workflow subagents can't prompt the user — that's why the asking happens here
in the main loop. Always relay the workflow's questions verbatim where useful.
- Keep accumulating answers across iterations; the workflow folds them into the
draft as settled decisions.
- Specs and tasks are created via `rulebook_*` MCP tools only.
Versions
| Repository | Score | Licence |
|---|---|---|
| This page · hivellm/rulebook | 90 | Apache-2.0 |
| renzrollon/interlock | 100 | MIT |
| CodaColaLight/nightmill | 100 | Apache-2.0 |
| duronext/jig | 100 | MIT |
| optiflowic/keyturn | 100 | MIT |
| rjmurillo/ai-agents | 100 | MIT |
| rjmurillo/ai-agents | 100 | MIT |
| rjmurillo/ai-agents | 100 | MIT |
| flavien-ia/hypervibe-craft | 100 | Apache-2.0 |
Identical content appears in 1 other GitHub folders. We show one copy with the highest score.
Database snapshot: 13.11.2026 · package version ba5760b33ce4 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100rulebook-sync
Use when rulebook/inbox.md has pending detections ("traiter l'inbox", "mettre à jour les règles"), when a hook or `rulebook check` reports that a rule "may have changed", "may have been removed" or "possible new rule", or when asked to sync, realign or update the rulebook after code changes. - 90rulebook-map
Use when a codebase's business rules are to be documented and no rulebook/index.md exists yet, or when asked to map, list or split the business domains of an application ("what are the domains", "cartographie des domaines", first step of a rulebook). - 100coograph-plan
Interview-driven planning. Investigates the codebase with code-graph, asks clarifying questions one at a time, and produces a work plan with acceptance criteria. Never implements. Use for new tickets or unclear requirements. - 91create-tasks
Generate Claude Code native Tasks from an existing spec. Use when user says "create tasks", "generate tasks from spec", "spec to tasks", "task generation", or wants to decompose a spec into implementation tasks. - 100absolute-spec
Lightweight standalone design spec for AI coding agents: codebase scan → bounded clarify pass (3–5 questions, not a grill) → reviewed design doc written to docs/plans/ → independent scored review → stop. No task board, no build. Use when you want a spec to discuss, hand off, or… - 91ansible-eda-rulebook
Create Event-Driven Ansible rulebooks and integrate with event sources for automated infrastructure response. Use this skill when the user asks to: "create eda rulebook", "event-driven ansible", "ansible rulebook", "eda event source", "webhook automation", "event automation",…