doc-driven-spec-workflow

Original🇺🇸 English
Translated

Use when a docs-driven repository request is unclear about whether the next step is scaffold initialization, clarification, roadmap decomposition, or current-task execution.

12installs
Added on

NPX Install

npx skill4agent add adol1111/doc-driven-spec-workflow doc-driven-spec-workflow

Using Doc-Driven Spec Workflow

Use this root skill to select and coordinate the right docs-driven workflow skill.
This skill is the entry protocol for the workflow. It does not own bootstrap, planning, specs, plans, implementation, templates, or branch-closing details. Those belong to the stage-specific skills.
This is analogous to
using-superpowers
: it makes the agent choose and respect the right composable skill instead of improvising the whole workflow from memory.

Mandatory Rules

  • MUST start docs-driven repository work by identifying the current workflow stage.
  • MUST load exactly one stage skill next unless the user explicitly asks for a comparison or audit.
  • MUST delegate stage behavior to the selected skill instead of restating or bypassing that skill's rules.
  • MUST treat roadmap planning and task reshaping as docs governance. They do not authorize spec writing or implementation by themselves.
  • MUST stop at approval gates.
    continue
    is not approval for roadmap structure, spec, plan, branch closing, or destructive cleanup.
  • MUST preserve handoff context when switching skills: what is decided, what is undecided, and why the next skill applies.

Workflow Chain

Default chain:
docs-workflow-bootstrap -> superpowers:brainstorming -> milestone-planning -> task-spec-execution
Not every request uses every stage. Skip stages only when their decision is already settled.

Routing Rules

Choose the next skill by the user's current uncertainty:
  • "We need the minimum docs scaffold first." ->
    docs-workflow-bootstrap
  • "We still do not understand what we are building." ->
    superpowers:brainstorming
  • "We understand the request, but not the milestone/module/task shape." ->
    milestone-planning
  • "We know the current concrete task and need spec-first execution." ->
    task-spec-execution
Use
docs-workflow-bootstrap
when the repository lacks:
  • docs/index.md
  • docs/architecture/index.md
  • docs/tasks/index.md
  • docs/context/index.md
Use
superpowers:brainstorming
when goals, constraints, success criteria, or "now versus later" boundaries are still ambiguous.
Use
milestone-planning
when the user needs milestone count, module grouping, task breakdown, delivery order, or stage boundaries.
Use
task-spec-execution
only after a concrete task exists or is selected and the user wants task-local
spec.md
, optional
plan.md
, readiness, implementation, verification, or branch closing.
Do not skip directly from ambiguous scope to spec writing. Do not use
task-spec-execution
to invent roadmap structure. Do not keep using
milestone-planning
after the question becomes current-task execution.

Handoff Rules

  • docs-workflow-bootstrap -> superpowers:brainstorming or milestone-planning
    : after the minimum docs scaffold exists
  • superpowers:brainstorming -> milestone-planning
    : after scope, boundaries, and success criteria are clear enough to plan delivery structure
  • milestone-planning -> task-spec-execution
    : after the current concrete task is selected and the user wants to enter spec-first execution
At each handoff, briefly restate:
  • what has already been decided
  • what remains undecided
  • what the next skill is responsible for
  • which explicit approval gate, if any, remains outstanding
Treat a user message such as
continue
as permission to advance to the next workflow stage only when the current stage is complete and no explicit approval gate is outstanding.
Do not treat
continue
as automatic approval for:
  • roadmap structure that still needs confirmation
  • spec approval
  • plan approval
  • branch closing or destructive cleanup decisions

Response Shape

Keep orchestration responses compact:
  1. Current stage
  2. Why this stage applies
  3. Next skill to use
  4. Expected stop point