Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Write a letter to the people you love, to be read later — the things you'd want them to know, the stories only you hold, the permission and the love that outlive you. Use when someone says 'help me write a letter to my kids/partner', 'legacy letter', 'ethical will', 'something for them to have when I'm gone', or is facing illness, aging, deployment, or simply wants to. Produces a warm, true letter in the writer's own voice — one to each person, or one to all — plus a light plan for when and how
.claude/skills/mohitagw15856-legacy-letter/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 104% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 52% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 69% | 0% |
| case-18 | ✗→✓ | ▲ Improved | 76% | 0% |
A will distributes your things; a legacy letter gives your people the things a will can't — the stories only you remember, what you were proud of, what you're sorry for, the specific reasons you loved them, and the permission to be okay. Also called an ethical will, it's one of the most treasured objects a family can hold, and most people mean to write one and never do because the blank page is too heavy. This skill makes it doable: it interviews gently, finds the true and specific things (not Hallmark platitudes), and shapes them into a letter that sounds unmistakably like the writer — because the voice is the gift.
their choice — built from true, specific memories and feelings, not generic sentiment
interview, usable in the letter and beyond
grieve and then live, the blessing on choices the writer won't be there for
the note that it can be revised anytime; this is a living document, not a final one
Ask for (if not already provided):
when they read it
gently, and only as much as they offer
they're proud of or sorry for, what they'd say if they had one more conversation
interview naturally — the skill mirrors their cadence)
is true and forgettable; "the way you refused to quit that ridiculous science project at 2am — that stubbornness is going to carry you" is unforgettable. The skill draws out concrete moments, exact sayings, the small true things. Specificity is what makes a reader feel seen years later.
including their humor, their bluntness, their particular words. A legacy letter that sounds like a template comforts no one; one that sounds like the person is like hearing them again. No inflated funeral-prose — earned warmth only.
an apology, a permission ("please don't stay sad — go live"), or a blessing on a future the writer won't see (marry them, take the job, be happy). The skill offers these gently and never forces them; it also never invents feelings the writer didn't express.
differs from the one they read at 30. If it's for a future moment, the skill can help write to who they'll be, or suggest a short set the writer adds to over time.
a tragedy. Light plan: where it's kept (with the will, told to a trusted person, or given directly), when it's meant to be read, and the standing invitation to revise it — people write these and live years more; it should grow with them.
## The letter(s)
[In the writer's own voice — one per person or one to all. True, specific, warm.
Hard-but-healing lines included only where the writer wanted them.]
## Stories we surfaced (yours to keep and reuse)
[The specific memories/sayings/lessons the interview found]
## When and how it's found
[Where it lives · when it's read · who's told · "revise it anytime — it's living"]"I'm so proud" filler standing alone
funeral speech
voice are the entire value
their letter, sacred and true; the skill is a scribe, not an author
assigning assets, gently route them to estate-planning-kit]]
or a confession
them, and it's fine to write it in pieces over time
digital-death-plan]] and estate-planning-kit]] for the practical legacy; grief-admin]] for the people who receive it; the-time-capsule]] is the professional-decisions cousin; personal-bio]] shares the voice-mining craft.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 27,232 | 23,532 | -14% | 1 | 1 | 0% | 3,005 | 3,865 | +29% | 0 | 0 | — |
case-02 | fail→fail | 15,433 | 18,567 | +20% | 1 | 1 | 0% | 2,349 | 3,463 | +47% | 0 | 0 | — |
case-03 | pass→pass | 18,075 | 13,696 | -24% | 1 | 1 | 0% | 2,264 | 2,715 | +20% | 0 | 0 | — |
case-04 | pass→pass | 20,791 | 24,798 | +19% | 1 | 1 | 0% | 2,664 | 4,488 | +68% | 0 | 0 | — |
case-05 | pass→pass | 21,981 | 25,156 | +14% | 1 | 1 | 0% | 2,746 | 4,500 | +64% | 0 | 0 | — |
case-06 | pass→pass | 16,492 | 19,380 | +18% | 1 | 1 | 0% | 1,801 | 3,492 | +94% | 0 | 0 | — |
case-07 | fail→fail | 14,492 | 21,284 | +47% | 1 | 1 | 0% | 1,659 | 3,675 | +122% | 0 | 0 | — |
case-08 | pass→pass | 17,442 | 10,018 | -43% | 1 | 1 | 0% | 1,839 | 2,832 | +54% | 0 | 0 | — |
case-09 | fail→pass | 20,817 | 15,173 | -27% | 1 | 1 | 0% | 2,324 | 3,637 | +56% | 0 | 0 | — |
case-10 | pass→pass | 21,012 | 17,379 | -17% | 1 | 1 | 0% | 2,474 | 3,269 | +32% | 0 | 0 | — |
case-11 | fail→pass | 9,701 | 10,661 | +10% | 1 | 1 | 0% | 1,544 | 3,147 | +104% | 0 | 0 | — |
case-12 | pass→pass | 12,314 | 17,125 | +39% | 1 | 1 | 0% | 1,882 | 3,752 | +99% | 0 | 0 | — |
case-13 | pass→pass | 17,764 | 23,709 | +33% | 1 | 1 | 0% | 2,781 | 4,081 | +47% | 0 | 0 | — |
case-14 | fail→fail | 17,230 | 15,806 | -8% | 1 | 1 | 0% | 1,870 | 3,074 | +64% | 0 | 0 | — |
case-15 | pass→pass | 11,422 | 16,541 | +45% | 1 | 1 | 0% | 1,872 | 3,165 | +69% | 0 | 0 | — |
case-21 | fail→pass | 19,763 | 16,315 | -17% | 1 | 1 | 0% | 2,008 | 3,059 | +52% | 0 | 0 | — |
case-16 | pass→pass | 5,590 | 8,942 | +60% | 1 | 1 | 0% | 1,030 | 2,866 | +178% | 0 | 0 | — |
case-17 | fail→pass | 12,972 | 11,928 | -8% | 1 | 1 | 0% | 1,513 | 2,558 | +69% | 0 | 0 | — |
case-18 | fail→pass | 14,917 | 14,810 | -1% | 1 | 1 | 0% | 1,729 | 3,043 | +76% | 0 | 0 | — |
case-19 | fail→fail | 21,729 | 18,556 | -15% | 1 | 1 | 0% | 2,344 | 3,495 | +49% | 0 | 0 | — |
case-20 | fail→fail | 19,532 | 24,956 | +28% | 1 | 1 | 0% | 2,352 | 3,947 | +68% | 0 | 0 | — |
case-22 | pass→pass | 18,346 | 14,556 | -21% | 1 | 1 | 0% | 1,899 | 3,431 | +81% | 0 | 0 | — |
case-23 | fail→pass | 17,429 | 18,245 | +5% | 1 | 1 | 0% | 2,003 | 3,478 | +74% | 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. 23 cases were attempted. The headline lift of +26 percentage points is the difference between those two pass rates over the 23 comparable cases. 1 case got worse with the skill loaded, and it is 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.