Writing a Design Spec

Take a design discussion and synthesize it into a written spec document. Run a fresh-eyes self-review, fix issues inline, gate on user approval, then recommend workbench:writing-plans as the next step.

When to Invoke

After a design discussion (typically a workbench:brainstorming session) where the user has approved the design verbally or in scratch notes. The discussion must be far enough along that the spec is a synthesis, not another round of questions.

If the user is still asking questions or exploring approaches, return them to workbench:brainstorming first.

Steps

You MUST create a task for each of these items and complete them in order:

1. Resolve the spec path based on profile or session context.
2. Draft the spec document at the resolved path.
3. Dispatch the self-review subagent with the prompt template in spec-document-reviewer-prompt.md.
4. Apply the reviewer's findings inline. No re-review.
5. Present to the user for approval.
6. Commit the spec if the path is committable.
7. Recommend `workbench:writing-plans` as the next skill.

Path Resolution

<paths.specs> resolves in order:

1. .workbench/autopilot.md Specs: heading.
2. Project CLAUDE.md or AGENTS.md spec-path convention.
3. .workbench/specs/ (default).

If the resolved value is don't commit, write the spec to /tmp/<project-name>-autopilot/YYYY-MM-DD-<topic>-design.md and skip the commit step. Otherwise write to <paths.specs>/YYYY-MM-DD-<topic>-design.md.

Drafting

Cover, scaling each section to its complexity:

• Problem.
• Solution overview.
• User stories (numbered).
• Implementation decisions (modules, interfaces, file layout, no code snippets).
• Testing decisions.
• Out of scope.
• Risks.

Use elements-of-style:writing-clearly-and-concisely if available.

Self-Review (Fresh-Eyes Subagent)

Dispatch a general-purpose subagent that has not seen the brainstorming conversation. The subagent reads the spec file and reports against four checks. You apply the findings; no re-review.

Claude Code: Agent tool, general-purpose subagent_type, no model override. Codex: equivalent general-purpose subagent.

The full reviewer prompt template lives in spec-document-reviewer-prompt.md in this skill's directory; paste it into the dispatch with [SPEC_FILE_PATH] replaced.

The four checks:

1. Placeholder scan: any "TBD", "TODO", incomplete sections, or vague requirements?
2. Internal consistency: do any sections contradict each other?
3. Scope check: is this focused enough for a single implementation plan?
4. Ambiguity check: could any requirement be interpreted two different ways?

After the report comes back, fix any issues inline in the spec document.

User Approval Gate

"Spec written and committed to <path>. Please review it and let me know if you want to make any changes before we start writing out the implementation plan."

Wait for the user's response. If they request changes, make them and re-run the self-review subagent. Only proceed to the next step once the user approves.

Handoff

The terminal action of this skill is recommending workbench:writing-plans. Do not invoke any implementation skill yourself. The spec is the artifact; the plan comes next.

Reference

spec-document-reviewer-prompt.md: full reviewer subagent prompt template.

Output Format

Default for this artifact: md.

Override resolution order, highest precedence first:

1. Per-invocation override in the user prompt. Recognize phrases like "an HTML spec", "in markdown", "as a markdown spec", "give me HTML", and equivalents.
2. .workbench/config.md ## Output formats entry for Specs:. Schema documented in plugins/workbench/skills/autopilot/references/config-schema.md.
3. Per-skill hard-coded default (md).

Path resolution order: .workbench/autopilot.md ## Documentation paths Specs: first; then .workbench/config.md ## Output paths Specs:; then .workbench/specs/ default. Path and format resolve independently.

When emitting HTML, follow the structural skeleton in references/spec-template.html in this skill's directory. Read the template lazily, only when actually producing the artifact. Do not introduce U+2014 or U+2013 codepoints in body copy; HTML entity forms (—, —, –, –) are permitted.

For other HTML artifact types not covered by a workbench or research skill, see workbench:crafting-html.

Applying a design system

Before emitting HTML, check for an active design system and inline its overrides into the artifact's <style> block:

1. Resolve the design-system name: per-prompt override (e.g., "render with the brand-2026 design system"), then .workbench/config.md ## Design system Name:, then no override.
2. Locate the directory: .workbench/design-systems/<name>/ (project scope), then ~/.claude/workbench/design-systems/<name>/ (user scope). If a name resolves but no directory is found at either scope, report the missing path to the user and emit with template defaults; do not fabricate a substitute.
3. Inline colors.css (and typography.css if present) after the template's own :root declarations, so the design system's values win the cascade.
4. For any referenced component, paste components/<n>.html markup and scoped style into the artifact body.
5. For any referenced image, base64-encode (base64 -w 0 <file>) and inline as data:image/<type>;base64,<payload>. SVG is text and can be inlined directly. Use relative paths only when the artifact and the design system co-exist in the same git tree and the artifact will not travel.

To create or edit a design system, see workbench:crafting-design-systems.