Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Direct-invocation specialist for clear, self-contained HTML plans that preserve source material while improving hierarchy, sequence, ownership, dependencies, and reviewability. Use when the user explicitly invokes html-plan or the broad html skill routes a plan request here. Do not activate independently from a general request.
.claude/skills/plannotator-html-plan/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 123% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -24% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 62% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -34% | 0% |
| case-04 | ✓→✗ | ▼ Worse | 20% | 0% |
Turn source material into a plan people can inspect and act on. Preserve the user's scope, ordering, commitments, and terminology unless they ask for broader synthesis.
Read the conversation, project instructions, and supplied plan before designing. Match an existing design language when one is present. Otherwise derive a quiet, workmanlike direction from the audience and subject.
When design-artifact is available, read its fundamentals to make that direction intentional. Keep this skill's traceability and source-preservation rules authoritative; creative direction must not inflate the plan into a dashboard or campaign page.
Decide what the plan actually needs:
Do not add a timeline, progress percentage, status badge, or dashboard summary unless the source supports it. Improve grammar and structure without inflating an implementation plan into a strategy document.
Deliver one responsive, accessible, self-contained HTML file. Use semantic headings, lists, tables, and landmarks. Keep essential CSS and JavaScript inline, avoid external services, and make any navigation or disclosure keyboard-operable.
Inspect the result at wide and narrow widths. Check that no commitment disappeared, that stages remain in the intended order, that ownership and dependencies are legible, and that long content does not overflow.
Return the absolute path and note any structural interpretation you introduced.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-07 | fail→pass | 11,049 | 24,368 | +121% | 1 | 1 | 0% | 2,955 | 6,603 | +123% | 0 | 0 | — |
case-01 | fail→pass | 26,344 | 23,152 | -12% | 1 | 1 | 0% | 6,212 | 4,713 | -24% | 0 | 0 | — |
case-08 | pass→fail | 19,191 | 10,602 | -45% | 1 | 1 | 0% | 4,481 | 2,954 | -34% | 0 | 0 | — |
case-02 | fail→fail | 26,140 | 26,438 | +1% | 1 | 1 | 0% | 6,074 | 6,611 | +9% | 0 | 0 | — |
case-03 | fail→fail | 27,531 | 30,626 | +11% | 1 | 1 | 0% | 6,197 | 6,604 | +7% | 0 | 0 | — |
case-04 | pass→fail | 23,201 | 28,501 | +23% | 1 | 1 | 0% | 5,482 | 6,583 | +20% | 0 | 0 | — |
case-05 | pass→fail | 19,551 | 27,989 | +43% | 1 | 1 | 0% | 4,564 | 6,580 | +44% | 0 | 0 | — |
case-06 | pass→fail | 23,164 | 27,985 | +21% | 1 | 1 | 0% | 3,852 | 6,579 | +71% | 0 | 0 | — |
case-09 | fail→fail | 25,764 | 13,480 | -48% | 1 | 1 | 0% | 5,881 | 2,693 | -54% | 0 | 0 | — |
case-10 | pass→fail | 8,727 | 14,292 | +64% | 1 | 1 | 0% | 2,077 | 2,407 | +16% | 0 | 0 | — |
case-11 | fail→fail | 34,570 | 21,659 | -37% | 1 | 1 | 0% | 3,530 | 5,205 | +47% | 0 | 0 | — |
case-12 | pass→fail | 21,180 | 13,217 | -38% | 1 | 1 | 0% | 2,711 | 3,656 | +35% | 0 | 0 | — |
case-13 | pass→fail | 21,195 | 11,072 | -48% | 1 | 1 | 0% | 5,362 | 3,020 | -44% | 0 | 0 | — |
case-14 | fail→fail | 25,013 | 27,349 | +9% | 1 | 1 | 0% | 6,174 | 6,580 | +7% | 0 | 0 | — |
case-15 | fail→pass | 17,927 | 26,087 | +46% | 1 | 1 | 0% | 4,070 | 6,579 | +62% | 0 | 0 | — |
case-16 | pass→fail | 17,446 | 27,135 | +56% | 1 | 1 | 0% | 3,479 | 6,575 | +89% | 0 | 0 | — |
case-17 | fail→fail | 7,854 | 73,222 | +832% | 1 | 1 | 0% | 1,496 | 6,578 | +340% | 0 | 0 | — |
case-18 | pass→fail | 8,270 | 6,087 | -26% | 1 | 1 | 0% | 1,888 | 1,330 | -30% | 0 | 0 | — |
case-19 | pass→pass | 5,269 | 22,157 | +321% | 1 | 1 | 0% | 1,165 | 5,867 | +404% | 0 | 0 | — |
case-20 | pass→pass | 6,013 | 25,710 | +328% | 1 | 1 | 0% | 1,234 | 6,592 | +434% | 0 | 0 | — |
case-21 | pass→pass | 15,213 | 5,340 | -65% | 1 | 1 | 0% | 3,556 | 1,305 | -63% | 0 | 0 | — |
case-22 | fail→fail | 6,417 | 16,887 | +163% | 1 | 1 | 0% | 960 | 3,486 | +263% | 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 -31 percentage points is the difference between those two pass rates over the 21 comparable cases. 9 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.