Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use after research (Z01 files exist) to create implementation plan - follow structured workflow
.claude/skills/nicepkg-feature-planning/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-13 | ✗→✓ | ▲ Improved | 173% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 64% | 0% |
| case-19 | ✓→✗ | ▼ Worse | 227% | 0% |
| case-12 | ✓→✗ | ▼ Worse | 179% | 0% |
| case-20 | ✓→✗ | ▼ Worse | 221% | 0% |
STOP. Before doing ANYTHING else:
in_progressThis skill is a WRAPPER that loads Z01 context and invokes superpowers:writing-plans
typescriptTodoWrite({ todos: [ {content: "Step 1: Load project context (CLAUDE.md if exists)", status: "in_progress", activeForm: "Reading CLAUDE.md"}, {content: "Step 2: Verify Z01 files exist", status: "pending", activeForm: "Checking research"}, {content: "Step 3: Read ALL Z01 files", status: "pending", activeForm: "Loading context"}, {content: "Step 4: Invoke superpowers:writing-plans", status: "pending", activeForm: "Creating plan"}, {content: "Step 5: Verify Z02 outputs", status: "pending", activeForm: "Validating output"} ] })
After each step: Mark completed, move in_progress to next step.
Without this wrapper: You'd manually load Z01 files, pass to superpowers:writing-plans, manage Z02 output paths, check for clarifications.
Read CLAUDE.md if it exists.
Extract from CLAUDE.md (if exists):
CRITICAL: If CLAUDE.md exists and contains constraints, these MUST be passed to superpowers:writing-plans so the plan preserves project standards.
Scan for existing Z01 files in common locations (docs/ai/ongoing, .ai/ongoing, docs/ongoing).
If Z01 files found:
If NO Z01 files found:
Read all Z01 files in ONGOING_DIR:
BLOCKING CHECK - If Z01_CLARIFY exists:
Extract:
CRITICAL: This skill's primary job is to invoke superpowers:writing-plans with Z01 context. If you skip this invocation, the skill provides no value.
Use Skill tool to invoke superpowers:writing-plans
Provide this instruction:
"Create an implementation plan for the {feature} feature based on the research in Z01_{feature}_research.md and clarifications in Z01_CLARIFY_{feature}_research.md.
MANDATORY CONSTRAINTS from CLAUDE.md: Include any constraints, patterns, or forbidden approaches from CLAUDE.md here if it exists]
CRITICAL: Save the plan to {ONGOING_DIR}/Z02_{feature}_plan.md (use the detected path, NOT hardcoded docs/plans/).
The plan should be a DIRECTIVE document with:
If you discover NEW blocking questions during planning (not already in Z01_CLARIFY), create {ONGOING_DIR}/Z02_CLARIFY_{feature}_plan.md. Otherwise, do NOT create a Z02_CLARIFY file.
When incorporating answered questions: Delete fully-answered CLARIFY files or remove incorporated Q&A pairs if only some were answered."
When superpowers:writing-plans completes:
Check structure:
Report to user:
/superpowers:write-plan (use Skill tool)| Excuse | Reality | |--------|---------| | "Skip path detection, I know it's docs/ai/ongoing" | NO. Path assumptions break in non-standard repos. Detect ONGOING_DIR. | | "No Z01 files, skip check" | NO. Research context critical for quality plans. Check first. | | "Superpowers will figure out output structure" | NO. Generic plans lack our research integration. Provide explicit Z02 instruction. | | "Read only Z01_research, skip Z01_CLARIFY" | NO. Missing context = incomplete plan. Read ALL Z01 files. | | "Create Z02_CLARIFY even if no questions" | NO. Empty files clutter directory. Only create if NEW questions. | | "Just invoke superpowers:writing-plans directly" | NO. This wrapper loads Z01 context. That's its value. | | "Wrapper skill, no need to track steps" | NO. Wrapper has critical steps (context loading, invocation). Track with TodoWrite. | | "TodoWrite adds overhead, skip it" | NO. TodoWrite provides user visibility and prevents skipped steps. MANDATORY. |
You followed the workflow if:
Workflow Position: AFTER feature-research (Z01 files), BEFORE feature-implement
Use when:
Don't use when:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→fail | 5,404 | 9,206 | +70% | 1 | 1 | 0% | 404 | 2,702 | +569% | 0 | 0 | — |
case-01 | fail→fail | 16,272 | 7,202 | -56% | 1 | 1 | 0% | 2,565 | 2,357 | -8% | 0 | 0 | — |
case-02 | fail→fail | 15,694 | 8,021 | -49% | 1 | 1 | 0% | 2,496 | 2,482 | -1% | 0 | 0 | — |
case-03 | fail→fail | 29,135 | 7,002 | -76% | 1 | 1 | 0% | 6,175 | 2,237 | -64% | 0 | 0 | — |
case-05 | fail→fail | 5,546 | 7,043 | +27% | 1 | 1 | 0% | 229 | 2,415 | +955% | 0 | 0 | — |
case-06 | fail→fail | 14,817 | 6,496 | -56% | 1 | 1 | 0% | 2,477 | 2,288 | -8% | 0 | 0 | — |
case-07 | fail→fail | 20,159 | 9,235 | -54% | 1 | 1 | 0% | 3,143 | 2,464 | -22% | 0 | 0 | — |
case-08 | fail→fail | 8,822 | 7,346 | -17% | 1 | 1 | 0% | 1,389 | 2,264 | +63% | 0 | 0 | — |
case-09 | fail→fail | 25,755 | 10,037 | -61% | 1 | 1 | 0% | 284 | 2,634 | +827% | 0 | 0 | — |
case-19 | pass→fail | 5,061 | 8,891 | +76% | 1 | 1 | 0% | 794 | 2,599 | +227% | 0 | 0 | — |
case-10 | pass→pass | 6,711 | 4,241 | -37% | 1 | 1 | 0% | 996 | 2,717 | +173% | 0 | 0 | — |
case-11 | fail→fail | 15,077 | 7,832 | -48% | 1 | 1 | 0% | 2,584 | 2,451 | -5% | 0 | 0 | — |
case-12 | pass→fail | 6,217 | 11,004 | +77% | 1 | 1 | 0% | 928 | 2,591 | +179% | 0 | 0 | — |
case-13 | fail→pass | 5,310 | 2,419 | -54% | 1 | 1 | 0% | 871 | 2,378 | +173% | 0 | 0 | — |
case-14 | pass→pass | 7,826 | 2,482 | -68% | 1 | 1 | 0% | 1,195 | 2,387 | +100% | 0 | 0 | — |
case-15 | fail→pass | 11,313 | 10,116 | -11% | 1 | 1 | 0% | 1,799 | 2,942 | +64% | 0 | 0 | — |
case-16 | fail→fail | 5,487 | 6,831 | +24% | 1 | 1 | 0% | 363 | 2,420 | +567% | 0 | 0 | — |
case-17 | fail→fail | 5,223 | 10,112 | +94% | 1 | 1 | 0% | 191 | 2,712 | +1320% | 0 | 0 | — |
case-18 | fail→fail | 17,064 | 7,980 | -53% | 1 | 1 | 0% | 2,606 | 2,346 | -10% | 0 | 0 | — |
case-20 | pass→fail | 5,750 | 10,121 | +76% | 1 | 1 | 0% | 889 | 2,854 | +221% | 0 | 0 | — |
case-21 | pass→fail | 9,150 | 12,643 | +38% | 1 | 1 | 0% | 1,510 | 2,735 | +81% | 0 | 0 | — |
case-22 | fail→fail | 9,280 | 8,721 | -6% | 1 | 1 | 0% | 1,460 | 2,306 | +58% | 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 4 counted toward the lift figure. The other 18 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 -9 percentage points is the difference between those two pass rates over the 4 comparable cases. 6 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.