Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Intent-driven knowhow precipitation — describe what you want to capture (记一个关于X的决策 / 保存这段代码模板 / 写个部署配方 / 存个调试技巧) and the workflow infers the type and records it into .workflow/knowhow/. Pure capture surface; knowhow 的管理/审计走 /maestro-knowledge；项目约束规则走 /maestro-spec add。Triggers on "knowhow capture", "知识沉淀", "沉淀经验", "记录模板", "记录决策", "adr", "存个技巧".
.claude/skills/catlog22-maestro-knowhow/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -35% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 0% | 0% |
<purpose> Intent-driven knowhow precipitation path (沉淀路径) — captures reusable knowledge into .workflow/knowhow/. No fixed grammar — state your intent; the knowhow step infers the content type and extracts the content. Type keywords still work as deterministic shortcuts:
| Type | Keywords | Prefix | |------|----------|--------| | session | session / compact / 压缩 | KNW- | | template | template / tpl / 模板 | TPL- | | recipe | recipe / rcp / 配方 / 步骤 | RCP- | | reference | reference / ref / 参考 | REF- | | decision | decision / dcs / adr / 决策 | DCS- | | tip | tip / note / 技巧 / 记录 | TIP- | | asset | asset / 资产 | AST- | | blueprint | blueprint / 蓝图 | BLP- | | document | document / doc / 文档 | DOC- | </purpose>
<dispatch> Read ~/.maestro/workflows/knowhow.md and follow the execution document directly. Do not create a Session or Run just to load this document.
Pass the full $ARGUMENTS to the workflow as its intent (the first capture argument is implied). The workflow infers one of the nine compatible types and writes through the canonical CLI.
Minimal ordinary creation is maestro knowhow add --type <type> --title "<title>" --content-file <path>: type, title, and content are the only required knowledge parameters. Keywords, sourceRef, relatedPaths, applicability, language, decision state, explicit ID, and tool are advanced optional metadata. Repository authority is host-owned: omit targetRepoId for current-repository writes; pass the exact host-supplied stable UUID only for an explicitly selected linked write with live knowhow write capability. Never derive identity from cwd/name/alias/path or persist those values as identity. </dispatch>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 46,066 | 17,732 | -62% | 1 | 1 | 0% | 1,882 | 866 | -54% | 0 | 0 | — |
case-02 | fail→fail | 36,833 | 16,757 | -55% | 1 | 1 | 0% | 1,126 | 724 | -36% | 0 | 0 | — |
case-03 | fail→fail | 11,048 | 12,477 | +13% | 1 | 1 | 0% | 751 | 876 | +17% | 0 | 0 | — |
case-04 | fail→pass | 17,486 | 9,601 | -45% | 1 | 1 | 0% | 1,596 | 1,037 | -35% | 0 | 0 | — |
case-05 | pass→pass | 23,056 | 3,995 | -83% | 1 | 1 | 0% | 1,042 | 997 | -4% | 0 | 0 | — |
case-06 | pass→pass | 9,330 | 5,026 | -46% | 1 | 1 | 0% | 1,268 | 1,123 | -11% | 0 | 0 | — |
case-07 | pass→pass | 23,591 | 9,914 | -58% | 1 | 1 | 0% | 1,100 | 947 | -14% | 0 | 0 | — |
case-08 | fail→fail | 14,589 | 35,633 | +144% | 1 | 1 | 0% | 1,495 | 827 | -45% | 0 | 0 | — |
case-09 | fail→pass | 15,701 | 15,419 | -2% | 1 | 1 | 0% | 1,875 | 1,046 | -44% | 0 | 0 | — |
case-10 | fail→pass | 18,713 | 4,615 | -75% | 1 | 1 | 0% | 1,734 | 943 | -46% | 0 | 0 | — |
case-11 | fail→pass | 13,447 | 10,280 | -24% | 1 | 1 | 0% | 1,372 | 1,143 | -17% | 0 | 0 | — |
case-12 | fail→pass | 20,455 | 9,476 | -54% | 1 | 1 | 0% | 938 | 937 | -0% | 0 | 0 | — |
case-13 | fail→pass | 15,845 | 8,952 | -44% | 1 | 1 | 0% | 1,519 | 982 | -35% | 0 | 0 | — |
case-14 | fail→pass | 15,647 | 11,348 | -27% | 1 | 1 | 0% | 1,289 | 745 | -42% | 0 | 0 | — |
case-15 | pass→pass | 11,202 | 7,626 | -32% | 1 | 1 | 0% | 1,866 | 736 | -61% | 0 | 0 | — |
case-16 | fail→pass | 19,413 | 11,271 | -42% | 1 | 1 | 0% | 2,403 | 797 | -67% | 0 | 0 | — |
case-17 | pass→pass | 14,981 | 7,582 | -49% | 1 | 1 | 0% | 1,268 | 724 | -43% | 0 | 0 | — |
case-18 | fail→pass | 15,825 | 7,321 | -54% | 1 | 1 | 0% | 1,405 | 685 | -51% | 0 | 0 | — |
case-19 | fail→pass | 15,878 | 2,131 | -87% | 1 | 1 | 0% | 1,450 | 594 | -59% | 0 | 0 | — |
case-20 | fail→pass | 16,868 | 5,767 | -66% | 1 | 1 | 0% | 1,501 | 697 | -54% | 0 | 0 | — |
case-21 | pass→pass | 16,221 | 8,160 | -50% | 1 | 1 | 0% | 1,588 | 750 | -53% | 0 | 0 | — |
case-22 | fail→pass | 16,075 | 7,994 | -50% | 1 | 1 | 0% | 1,603 | 703 | -56% | 0 | 0 | — |
case-23 | fail→pass | 5,222 | 10,047 | +92% | 1 | 1 | 0% | 675 | 1,179 | +75% | 0 | 0 | — |
case-24 | fail→pass | 12,596 | 11,001 | -13% | 1 | 1 | 0% | 1,176 | 1,465 | +25% | 0 | 0 | — |
case-25 | fail→pass | 11,747 | 8,839 | -25% | 1 | 1 | 0% | 967 | 941 | -3% | 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. 25 cases were attempted, and 21 counted toward the lift figure. The other 4 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 +60 percentage points is the difference between those two pass rates over the 21 comparable cases.
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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/30/2026 | — |
| gemini-3.6-flash | verified | 8/21/2026 | +17% |
Other measured skills in the registry, with their headline benchmark lift.