Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →This skill should be used when the user asks to "build a feature", "create a spec", "start spec-driven development", "run research phase", "generate requirements", "create design", "plan tasks", "implement spec", "check spec status", "triage a feature", "create an epic", "decompose a large feature", or needs guidance on spec-driven development workflow, phase ordering, or epic orchestration.
.claude/skills/tzachbon-spec-workflow/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 8% | 0% |
Spec-driven development transforms feature requests into structured specs through sequential phases, then executes them task-by-task.
| Situation | Command | |-----------|---------| | New feature, want guidance | /ralph-specum:start <name> <goal> | | New feature, skip interviews | /ralph-specum:start <name> <goal> --quick | | Large feature needing decomposition | /ralph-specum:triage <goal> | | Resume existing spec | /ralph-specum:start (auto-detects) | | Jump to specific phase | /ralph-specum:<phase> |
start/new -> research -> requirements -> design -> tasks -> implementEach phase produces a markdown artifact in ./specs/<name>/. Normal mode pauses for approval between phases. Quick mode runs all phases then auto-starts execution.
| Command | Agent | Output | Purpose | |---------|-------|--------|---------| | /ralph-specum:research | research-analyst | research.md | Explore feasibility, patterns, context | | /ralph-specum:requirements | product-manager | requirements.md | User stories, acceptance criteria | | /ralph-specum:design | architect-reviewer | design.md | Architecture, components, interfaces | | /ralph-specum:tasks | task-planner | tasks.md | POC-first task breakdown | | /ralph-specum:implement | spec-executor | commits | Autonomous task-by-task execution |
For features too large for a single spec, use epic triage to decompose into dependency-aware specs.
triage -> [spec-1, spec-2, spec-3...] -> implement each in orderEntry points:
/ralph-specum:triage <goal> -- create or resume an epic/ralph-specum:start -- detects active epics, suggests next unblocked specFile structure:
specs/
_epics/<epic-name>/
epic.md # Triage output (vision, specs, dependency graph)
research.md # Exploration + validation research
.epic-state.json # Progress tracking across specs
.progress.md # Learnings and decisions| Command | Purpose | |---------|---------| | /ralph-specum:status | Show all specs and progress | | /ralph-specum:switch <name> | Change active spec | | /ralph-specum:cancel | Cancel active execution | | /ralph-specum:refactor | Update spec files after execution |
bash/ralph-specum:start my-feature "Build X" --quick # Runs all phases automatically, starts execution
bash/ralph-specum:start my-feature "Build X" # Interactive interviews at each phase # Review and approve each artifact /ralph-specum:implement
bash/ralph-specum:triage "Build entire auth system" # Decomposes into: auth-core, auth-oauth, auth-rbac /ralph-specum:start # Picks next unblocked spec
references/phase-transitions.md -- Detailed phase flow, state transitions, quick mode behavior, phase skipping| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 15,878 | 2,622 | -83% | 1 | 1 | 0% | 2,651 | 1,221 | -54% | 0 | 0 | — |
case-02 | fail→pass | 17,821 | 4,330 | -76% | 1 | 1 | 0% | 2,762 | 1,529 | -45% | 0 | 0 | — |
case-03 | fail→pass | 11,022 | 4,302 | -61% | 1 | 1 | 0% | 1,582 | 1,495 | -5% | 0 | 0 | — |
case-04 | fail→pass | 9,199 | 2,147 | -77% | 1 | 1 | 0% | 1,389 | 1,122 | -19% | 0 | 0 | — |
case-05 | fail→pass | 6,349 | 1,884 | -70% | 1 | 1 | 0% | 994 | 1,073 | +8% | 0 | 0 | — |
case-06 | fail→pass | 6,035 | 1,660 | -72% | 1 | 1 | 0% | 907 | 986 | +9% | 0 | 0 | — |
case-07 | pass→pass | 15,580 | 1,621 | -90% | 1 | 1 | 0% | 2,521 | 1,050 | -58% | 0 | 0 | — |
case-08 | fail→pass | 8,748 | 1,515 | -83% | 1 | 1 | 0% | 1,247 | 981 | -21% | 0 | 0 | — |
case-09 | fail→pass | 7,432 | 1,806 | -76% | 1 | 1 | 0% | 1,103 | 1,047 | -5% | 0 | 0 | — |
case-10 | fail→pass | 6,600 | 2,222 | -66% | 1 | 1 | 0% | 1,058 | 1,175 | +11% | 0 | 0 | — |
case-11 | fail→pass | 9,212 | 2,235 | -76% | 1 | 1 | 0% | 1,409 | 1,168 | -17% | 0 | 0 | — |
case-12 | fail→pass | 10,834 | 4,547 | -58% | 1 | 1 | 0% | 1,475 | 1,542 | +5% | 0 | 0 | — |
case-13 | fail→pass | 17,735 | 2,582 | -85% | 1 | 1 | 0% | 1,045 | 1,189 | +14% | 0 | 0 | — |
case-14 | fail→pass | 12,101 | 1,779 | -85% | 1 | 1 | 0% | 1,885 | 1,031 | -45% | 0 | 0 | — |
case-15 | fail→pass | 10,740 | 1,959 | -82% | 1 | 1 | 0% | 1,707 | 1,050 | -38% | 0 | 0 | — |
case-16 | pass→pass | 9,078 | 1,761 | -81% | 1 | 1 | 0% | 1,297 | 1,039 | -20% | 0 | 0 | — |
case-17 | pass→pass | 9,128 | 2,505 | -73% | 1 | 1 | 0% | 1,437 | 1,216 | -15% | 0 | 0 | — |
case-18 | fail→pass | 16,676 | 3,513 | -79% | 1 | 1 | 0% | 2,236 | 1,253 | -44% | 0 | 0 | — |
case-19 | fail→pass | 11,511 | 2,150 | -81% | 1 | 1 | 0% | 1,537 | 1,138 | -26% | 0 | 0 | — |
case-20 | pass→pass | 5,268 | 6,510 | +24% | 1 | 1 | 0% | 1,037 | 2,128 | +105% | 0 | 0 | — |
case-21 | pass→pass | 5,706 | 3,755 | -34% | 1 | 1 | 0% | 912 | 1,385 | +52% | 0 | 0 | — |
case-22 | pass→pass | 16,686 | 18,607 | +12% | 1 | 1 | 0% | 2,588 | 3,609 | +39% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +73 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.