Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Free lite skill — drafts one friendly (level-1) client document-request email for a single tax client from a short brief plus the practice profile. A strict subset of client-doc-chaser (no batch, no escalation levels 2/3), carrying the pack's full guard. Use to try the kit before buying, or as the directory-published funnel skill. Triggers - "client document request email", "chase a client's documents", "free document request writer", "doc-chaser-lite".
.claude/skills/davila7-doc-chaser-lite/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 165% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 383% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 16% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 289% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 240% | 0% |
The free lite subset of client-doc-chaser. It drafts exactly one friendly, level-1 document-request email in the firm's own voice, for one client, at professional tone. It drafts only: it never sends, never computes or estimates a figure, never invents a date, and never gives advice. It carries the pack's full guard — the guard is never the paywall. It deliberately does less than the paid skill: no batch runs, no escalation levels 2 or 3, no consequence lines.
ABOUT-ME/profile.md (required): firm/practice name; preparer name + credential;contact block (email + phone — the sign-off renders from it); services offered. Tone is fixed at professional for the lite skill (the warm preset is a full-kit feature — the tone field in the profile is ignored here).
brief-template.md, carrying: clientinitials or case ID; engagement type; outstanding-document list; the accountant-stated due date. There is no escalation field (every lite email is a friendly level-1 nudge) and no consequence line.
consequence, and the lite skill never invents one or firms up the tone to imitate a paid level.
explainer, an onboarding packet, or a deadline pack, the lite skill still delivers the one level-1 email it can and names the full kit for the rest (see the upsell footer). It never half-delivers a paid feature and never fakes one (US6 sc.3).
self-reported number (counts are checked externally, not by the model).
A single draft (the lite skill never batches), in this order:
--- COPY ABOVE THIS LINE ---.accuracy and for the professional standards that apply to your practice. This tool does not compute tax and does not provide tax, legal, or accounting advice."
(never above the --- COPY ABOVE THIS LINE --- line), never more than once per generation. One line naming the full kit and the utm_source=lite-footer tracked Gumroad URL:
> Want batch mode, all three escalation levels, and the full Accountant Cowork Kit > (engagement letters, notice explainers, onboarding packets, deadline reminders)? Get it at > https://moailoops.gumroad.com/l/accountant-cowork-kit?utm_source=lite-footer
<!-- FR-083: the surface-tagged ?utm_source=lite-footer link is finalized (T19). The base Gumroad URL (host + product slug) is confirmed when the listing goes live (T16); if it differs from the placeholder above, update this footer, the PROMPT-ONLY.md copy, and the golden example in one pass. Keep the footer to a single line, after the separator, never inside the draft. -->
Some directories ban promotional content inside skill output. The variant published to such a directory omits the upsell footer entirely (drop output-structure item 5) and moves the kit mention to the directory listing description and this skill's README. This is a deletion-only change — Tier 3 exempt. The bundle-zip copy always keeps the footer; only a directory build that is contractually barred from in-output promotion drops it.
wait. Never invent a date, figure, or document, and never fill a gap from memory (never more than one clarify round per generation).
guard.md as a final self-audit pass before returning output — it is the single source oftruth for what this skill may and may not do; this file does not re-derive its rules in its own words. The lite skill carries the guard in full — nothing in it is relaxed because it is free.
See examples/doc-chaser-lite-golden.md in this folder for the worked example (the golden level-1 email plus its footer). This skill ships with the same guard and passes the same adversarial red-team suite as the paid kit before every release — the guard is never the paywall.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | pass→fail | 4,320 | 3,872 | -10% | 1 | 1 | 0% | 907 | 2,024 | +123% | 0 | 0 | — |
case-01 | fail→pass | 8,826 | 14,082 | +60% | 1 | 1 | 0% | 1,626 | 4,312 | +165% | 0 | 0 | — |
case-02 | fail→pass | 5,364 | 17,011 | +217% | 1 | 1 | 0% | 1,001 | 4,836 | +383% | 0 | 0 | — |
case-03 | fail→fail | 7,393 | 11,496 | +55% | 1 | 1 | 0% | 1,334 | 3,591 | +169% | 0 | 0 | — |
case-04 | fail→fail | 9,303 | 7,086 | -24% | 1 | 1 | 0% | 1,667 | 2,636 | +58% | 0 | 0 | — |
case-05 | fail→fail | 8,793 | 7,354 | -16% | 1 | 1 | 0% | 1,525 | 2,677 | +76% | 0 | 0 | — |
case-06 | fail→fail | 8,425 | 7,341 | -13% | 1 | 1 | 0% | 1,474 | 2,711 | +84% | 0 | 0 | — |
case-07 | fail→fail | 21,384 | 6,763 | -68% | 1 | 1 | 0% | 3,897 | 2,682 | -31% | 0 | 0 | — |
case-08 | fail→fail | 7,133 | 5,493 | -23% | 1 | 1 | 0% | 1,232 | 2,307 | +87% | 0 | 0 | — |
case-10 | fail→fail | 5,323 | 3,882 | -27% | 1 | 1 | 0% | 1,100 | 2,071 | +88% | 0 | 0 | — |
case-11 | fail→fail | 4,088 | 2,937 | -28% | 1 | 1 | 0% | 775 | 1,880 | +143% | 0 | 0 | — |
case-12 | pass→pass | 4,616 | 12,567 | +172% | 1 | 1 | 0% | 850 | 3,872 | +356% | 0 | 0 | — |
case-13 | fail→fail | 10,450 | 5,177 | -50% | 1 | 1 | 0% | 2,048 | 2,286 | +12% | 0 | 0 | — |
case-14 | fail→pass | 14,126 | 7,229 | -49% | 1 | 1 | 0% | 2,374 | 2,763 | +16% | 0 | 0 | — |
case-15 | pass→pass | 16,368 | 6,119 | -63% | 1 | 1 | 0% | 3,214 | 2,647 | -18% | 0 | 0 | — |
case-16 | fail→pass | 4,355 | 10,062 | +131% | 1 | 1 | 0% | 865 | 3,363 | +289% | 0 | 0 | — |
case-17 | fail→pass | 7,621 | 15,829 | +108% | 1 | 1 | 0% | 1,430 | 4,858 | +240% | 0 | 0 | — |
case-18 | fail→pass | 9,976 | 4,892 | -51% | 1 | 1 | 0% | 1,858 | 2,139 | +15% | 0 | 0 | — |
case-19 | fail→fail | 5,072 | 7,686 | +52% | 1 | 1 | 0% | 988 | 3,045 | +208% | 0 | 0 | — |
case-20 | pass→fail | 8,690 | 4,077 | -53% | 1 | 1 | 0% | 2,129 | 2,035 | -4% | 0 | 0 | — |
case-21 | pass→fail | 10,943 | 5,594 | -49% | 1 | 1 | 0% | 2,613 | 2,321 | -11% | 0 | 0 | — |
case-22 | pass→fail | 13,268 | 6,525 | -51% | 1 | 1 | 0% | 2,845 | 2,660 | -7% | 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. The headline lift of +9 percentage points is the difference between those two pass rates over the 22 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.