From dev-flow
Methodology reference for generating implementation specs. Invoked via /df:spec command.
How this skill is triggered — by the user, by Claude, or both
Slash command
/dev-flow:specThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Transform triage output and user decisions into a detailed implementation specification ready for development.
Transform triage output and user decisions into a detailed implementation specification ready for development.
The skill expects one of:
.aiwork/*/triage.md (any task folder).aiwork/*<ticket-id>*/triage.md, or fetch from Azure DevOpsExtract from triage:
Use AskUserQuestion tool to gather decisions. Organize questions into batches of 3-4 max.
Standard question categories:
Mark questions that need team input (can't be answered by assignee alone).
If not already done in triage, use Explore agent to gather:
Output location: If the project defines an .aiwork/ folder protocol (e.g., naming conventions, frontmatter, folder structure), follow that protocol. Otherwise use these defaults:
Write to .aiwork/{date}_{slug}/spec.md
Where:
{date} = current date as YYYY-MM-DD{slug} = with ticket ID: <ticket-id>-<slugified-title>, without: <slugified-title>If a task folder already exists (from triage or otherwise), place spec.md in the same folder.
Use the template from ${CLAUDE_SKILL_DIR}/assets/spec-template.md.
"Where should <component> live after migration?"
Options:
- "<layer-a> (Recommended)" - <rationale>
- "<layer-b>" - <rationale>
- "Keep in current location" - <rationale>
"Should <dependency> be completed first, or included in this task?"
Options:
- "Wait for <ticket>" - Block until dependency completes
- "Include in this task" - Expand scope to include it
- "Parallel work OK" - Proceed independently, coordinate later
"<Technology> approach?"
Options:
- "<Option A> (Recommended)" - <rationale>
- "<Option B>" - <rationale>
npx claudepluginhub lttr/claude-marketplace --plugin dfConducts multi-round interviews to refine rough SPEC.md into complete, implementation-ready specifications with tasks. Use for new features, requirements refinement, or ideas to actionable specs.
Creates technical specifications interactively by gathering requirements, exploring codebases, running planning interviews, drafting with Mermaid diagrams, expert review, and iteration. For new features or projects needing architecture and decisions.
Generates dynamic phased implementation plans from approved specs for execution coordination, agent assignments, and dependencies in /sdd workflows.