Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Execute an implementation plan. Implements every task, runs validation, commits, and opens a PR. Use when the user asks to execute, implement, or run a plan file.
.claude/skills/itsbencodes-execute-plan/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -43% | 0% |
Implement a plan end-to-end: code the tasks, validate, commit, and open a PR.
Read the plan file passed in $ARGUMENTS. If no path is given, list .agents/plans/ and ask which one. Re-read the plan whenever you need to refresh context — do not paraphrase from memory.
Never commit to main. Check the current branch:
main, create a new branch named after the plan slug (e.g., <slug> from YYYY-MM-DD-<slug>.md).Use TaskCreate to mirror every item from the plan's "Tasks" section as a todo, in order. The final todo must be "Run validation steps" (this should already be the last task in the plan). Do not skip, merge, or reorder tasks.
Work through todos one at a time. Mark each completed as soon as it is done — do not batch. If you discover the plan is wrong or incomplete, stop and ask the user before diverging.
Use the validate skill.
Use the commit skill. Stage only plan-related files. Default base is main (see memory).
One or two sentences: what shipped, PR link, anything the user should know (skipped items, open questions, follow-ups).
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,018 | 15,420 | +156% | 1 | 1 | 0% | 337 | 772 | +129% | 0 | 0 | — |
case-02 | fail→fail | 5,646 | 4,524 | -20% | 1 | 1 | 0% | 279 | 553 | +98% | 0 | 0 | — |
case-03 | fail→fail | 5,153 | 6,382 | +24% | 1 | 1 | 0% | 313 | 683 | +118% | 0 | 0 | — |
case-04 | fail→fail | 3,562 | 5,277 | +48% | 1 | 1 | 0% | 517 | 599 | +16% | 0 | 0 | — |
case-05 | fail→fail | 6,577 | 6,783 | +3% | 1 | 1 | 0% | 313 | 752 | +140% | 0 | 0 | — |
case-06 | pass→pass | 12,913 | 1,933 | -85% | 1 | 1 | 0% | 2,110 | 649 | -69% | 0 | 0 | — |
case-07 | fail→fail | 9,011 | 9,049 | +0% | 1 | 1 | 0% | 1,402 | 1,041 | -26% | 0 | 0 | — |
case-08 | pass→fail | 8,313 | 2,625 | -68% | 1 | 1 | 0% | 1,279 | 729 | -43% | 0 | 0 | — |
case-09 | fail→pass | 3,655 | 3,017 | -17% | 1 | 1 | 0% | 710 | 899 | +27% | 0 | 0 | — |
case-10 | fail→fail | 4,619 | 28,524 | +518% | 1 | 1 | 0% | 789 | 841 | +7% | 0 | 0 | — |
case-11 | fail→fail | 5,423 | 13,797 | +154% | 1 | 1 | 0% | 755 | 1,789 | +137% | 0 | 0 | — |
case-12 | pass→pass | 4,450 | 2,379 | -47% | 1 | 1 | 0% | 579 | 729 | +26% | 0 | 0 | — |
case-13 | fail→fail | 12,324 | 23,290 | +89% | 1 | 1 | 0% | 1,889 | 3,696 | +96% | 0 | 0 | — |
case-14 | fail→fail | 2,715 | 1,830 | -33% | 1 | 1 | 0% | 457 | 615 | +35% | 0 | 0 | — |
case-15 | fail→pass | 8,329 | 1,550 | -81% | 1 | 1 | 0% | 1,368 | 520 | -62% | 0 | 0 | — |
case-16 | pass→pass | 7,619 | 2,778 | -64% | 1 | 1 | 0% | 1,057 | 720 | -32% | 0 | 0 | — |
case-17 | fail→pass | 9,106 | 2,358 | -74% | 1 | 1 | 0% | 1,335 | 631 | -53% | 0 | 0 | — |
case-18 | fail→pass | 10,352 | 5,470 | -47% | 1 | 1 | 0% | 950 | 1,056 | +11% | 0 | 0 | — |
case-19 | pass→pass | 8,114 | 3,614 | -55% | 1 | 1 | 0% | 1,400 | 1,029 | -27% | 0 | 0 | — |
case-20 | pass→fail | 8,005 | 5,879 | -27% | 1 | 1 | 0% | 1,150 | 825 | -28% | 0 | 0 | — |
case-21 | fail→fail | 12,766 | 5,396 | -58% | 1 | 1 | 0% | 386 | 653 | +69% | 0 | 0 | — |
case-22 | pass→fail | 14,312 | 6,281 | -56% | 1 | 1 | 0% | 2,976 | 527 | -82% | 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 13 counted toward the lift figure. The other 9 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 +5 percentage points is the difference between those two pass rates over the 13 comparable cases. 5 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.