Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Plan the next PRD phase by drafting issues, creating GitHub issues, and bridging planning to building.
.claude/skills/joshukraine-plan-phase/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -19% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -17% | 0% |
Plan the next phase of development from the PRD, create GitHub issues, and prepare for implementation. This skill bridges planning and building — it creates the structure but writes no application code.
Work through the steps below in order. Each step has a hard stop — do not proceed past a CHECKPOINT without explicit user approval.
> Tip for the user: You can activate Claude Code's plan mode (shift+tab) during Steps 1-2 for an extra guardrail against premature code changes. Toggle out of plan mode before Step 3 so that Claude Code can run gh issue create commands.
docs/prd/ROADMAP.md to identify the current phase and what comes nextdocs/prd/*-overview.md) for project contextCLAUDE.mdPresent a clear, concise plan for the next phase:
02-auth-and-onboarding.md")feat: ... or fix: ... or chore: ...Keep the number of issues manageable — typically 3-7 per phase. If the phase is larger, break it into sub-phases.
CHECKPOINT 1: Present the plan and wait for user approval. The user may adjust scope, reorder issues, split or combine items, or request changes. Do not proceed until the user explicitly approves.
After plan approval, create each issue on GitHub using the gh CLI:
bashgh issue create --title "feat: [title]" --body "[description]"
Each issue body should include:
02-auth-and-onboarding.md")Create the issues in implementation order. Record each issue number as it is created.
After all issues are created, present a summary:
textPhase: [name] Issues created: #41 - feat: Add User model with Devise authentication (medium) #42 - feat: Add Pundit authorization policies (small) #43 - feat: Add admin dashboard with user management (large) Recommended starting point: /resolve-issue 41
CHECKPOINT 2: Present this summary and wait for user approval before any implementation begins. Do not write application code, create branches, or modify any files beyond GitHub issues.
When the user is ready to begin, suggest the /resolve-issue command for the first issue in sequence.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-19 | fail→fail | 7,199 | 1,946 | -73% | 1 | 1 | 0% | 1,073 | 1,183 | +10% | 0 | 0 | — |
case-01 | fail→fail | 1,981 | 4,008 | +102% | 1 | 1 | 0% | 282 | 1,163 | +312% | 0 | 0 | — |
case-02 | fail→fail | 4,821 | 5,474 | +14% | 1 | 1 | 0% | 728 | 1,171 | +61% | 0 | 0 | — |
case-03 | fail→pass | 10,575 | 6,528 | -38% | 1 | 1 | 0% | 2,452 | 1,903 | -22% | 0 | 0 | — |
case-04 | fail→fail | 23,558 | 4,574 | -81% | 1 | 1 | 0% | 3,979 | 1,186 | -70% | 0 | 0 | — |
case-05 | pass→fail | 9,661 | 7,216 | -25% | 1 | 1 | 0% | 1,982 | 1,191 | -40% | 0 | 0 | — |
case-06 | fail→fail | 9,244 | 3,854 | -58% | 1 | 1 | 0% | 1,439 | 1,349 | -6% | 0 | 0 | — |
case-07 | pass→pass | 11,343 | 5,238 | -54% | 1 | 1 | 0% | 1,811 | 1,700 | -6% | 0 | 0 | — |
case-08 | pass→pass | 12,711 | 5,951 | -53% | 1 | 1 | 0% | 2,013 | 1,754 | -13% | 0 | 0 | — |
case-09 | pass→pass | 13,271 | 3,660 | -72% | 1 | 1 | 0% | 1,969 | 1,409 | -28% | 0 | 0 | — |
case-10 | fail→fail | 13,750 | 6,414 | -53% | 1 | 1 | 0% | 2,135 | 1,322 | -38% | 0 | 0 | — |
case-11 | pass→pass | 7,960 | 3,967 | -50% | 1 | 1 | 0% | 1,338 | 1,432 | +7% | 0 | 0 | — |
case-12 | fail→pass | 12,102 | 4,758 | -61% | 1 | 1 | 0% | 1,979 | 1,652 | -17% | 0 | 0 | — |
case-13 | pass→fail | 8,296 | 2,198 | -74% | 1 | 1 | 0% | 1,204 | 1,198 | -0% | 0 | 0 | — |
case-14 | fail→pass | 11,339 | 3,957 | -65% | 1 | 1 | 0% | 1,889 | 1,535 | -19% | 0 | 0 | — |
case-15 | fail→pass | 8,381 | 1,852 | -78% | 1 | 1 | 0% | 1,203 | 1,228 | +2% | 0 | 0 | — |
case-16 | pass→fail | 5,324 | 7,100 | +33% | 1 | 1 | 0% | 845 | 1,281 | +52% | 0 | 0 | — |
case-17 | pass→fail | 6,810 | 3,163 | -54% | 1 | 1 | 0% | 965 | 1,145 | +19% | 0 | 0 | — |
case-18 | fail→pass | 11,510 | 3,799 | -67% | 1 | 1 | 0% | 1,726 | 1,427 | -17% | 0 | 0 | — |
case-20 | fail→pass | 9,321 | 3,358 | -64% | 1 | 1 | 0% | 1,373 | 1,335 | -3% | 0 | 0 | — |
case-21 | pass→pass | 10,547 | 6,173 | -41% | 1 | 1 | 0% | 1,725 | 1,770 | +3% | 0 | 0 | — |
case-22 | fail→pass | 10,068 | 2,836 | -72% | 1 | 1 | 0% | 1,571 | 1,366 | -13% | 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 16 counted toward the lift figure. The other 6 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 +14 percentage points is the difference between those two pass rates over the 16 comparable cases. 4 cases got worse with the skill loaded, and they are included in that figure.
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.