Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Scaffold a custom AI agent harness via `metaharness new <name> --template <id> --host <id>`. Defaults to DRY-RUN (no writes) unless --confirm is passed. Refuses to write to the calling repo root or anywhere inside it. Honors ADR-150 architectural constraint + ruflo's "destructive-action confirmation" pattern.
.claude/skills/ruvnet-harness-mint/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-15 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -34% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 53% | 0% |
The one write-capable skill in the plugin. Every other skill is pure-read. This one calls metaharness new, which writes a new directory tree.
--confirm, the script prints whatit would do and exits 0 without touching disk.
--target resolves to the currentworking directory OR any path inside it, the script errors out with exit 2. Target must be an absolute path OUTSIDE the calling repo (default is a fresh /tmp/ruflo-mint-<ts>-<name>/ dir).
non-existent dir.
execution. The mint stays sandboxed from ruflo's runtime.
Implementation: scripts/mint.mjs.
--name, --template. Default --host to claude-code.--target (default: temp dir).--confirm: emit dry-run plan, exit 0.--confirm: shell npx metaharness new <name> --template <id>--host <id> --target <abs> --yes.
minimal, vertical:coding, vertical:devops, vertical:support, vertical:legal, vertical:research, vertical:trading, vertical:health, vertical:education, vertical:sales, vertical:business, vertical:crm, vertical:marketing, vertical:advertising, vertical:ai, vertical:agentics, vertical:ruview, vertical:gaming, vertical:repo-maintainer, vertical:exotic.
claude-code, codex, pi-dev, hermes, openclaw, rvm, copilot, opencode, github-actions.
$ node scripts/mint.mjs --name my-harness --template vertical:coding --host claude-code
# harness-mint (dry-run)
- action: metaharness new
- name: my-harness
- template: vertical:coding
- host: claude-code
- target: /tmp/ruflo-mint-1718560000-my-harness
- confirm: false
- willWrite: false
Re-run with `--confirm` to actually scaffold.Ruflo's behavioral rules say "executing actions with care" — destructive or repo-touching actions need confirmation. The dry-run output makes the WHAT visible before the WHEN. A human sees target, decides, then adds --confirm if happy.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-15 | fail→pass | 6,931 | 2,695 | -61% | 1 | 1 | 0% | 1,260 | 1,177 | -7% | 0 | 0 | — |
case-01 | fail→pass | 9,588 | 6,763 | -29% | 1 | 1 | 0% | 1,623 | 1,989 | +23% | 0 | 0 | — |
case-02 | fail→pass | 7,635 | 1,858 | -76% | 1 | 1 | 0% | 1,504 | 994 | -34% | 0 | 0 | — |
case-03 | fail→pass | 9,772 | 4,952 | -49% | 1 | 1 | 0% | 2,098 | 1,623 | -23% | 0 | 0 | — |
case-04 | fail→pass | 6,421 | 4,859 | -24% | 1 | 1 | 0% | 1,083 | 1,658 | +53% | 0 | 0 | — |
case-05 | fail→pass | 10,457 | 966 | -91% | 1 | 1 | 0% | 1,857 | 829 | -55% | 0 | 0 | — |
case-06 | fail→pass | 7,201 | 3,093 | -57% | 1 | 1 | 0% | 1,129 | 1,173 | +4% | 0 | 0 | — |
case-07 | fail→pass | 11,690 | 1,973 | -83% | 1 | 1 | 0% | 1,111 | 1,033 | -7% | 0 | 0 | — |
case-08 | fail→pass | 20,700 | 1,272 | -94% | 1 | 1 | 0% | 1,824 | 889 | -51% | 0 | 0 | — |
case-09 | fail→pass | 15,547 | 2,660 | -83% | 1 | 1 | 0% | 1,375 | 1,208 | -12% | 0 | 0 | — |
case-10 | fail→pass | 26,470 | 3,905 | -85% | 1 | 1 | 0% | 1,087 | 1,474 | +36% | 0 | 0 | — |
case-11 | fail→pass | 10,165 | 4,335 | -57% | 1 | 1 | 0% | 1,748 | 1,479 | -15% | 0 | 0 | — |
case-12 | fail→pass | 18,029 | 1,827 | -90% | 1 | 1 | 0% | 960 | 977 | +2% | 0 | 0 | — |
case-13 | pass→pass | 6,802 | 1,277 | -81% | 1 | 1 | 0% | 1,162 | 885 | -24% | 0 | 0 | — |
case-14 | fail→pass | 8,371 | 1,921 | -77% | 1 | 1 | 0% | 1,588 | 1,008 | -37% | 0 | 0 | — |
case-16 | fail→pass | 10,902 | 2,762 | -75% | 1 | 1 | 0% | 2,008 | 1,161 | -42% | 0 | 0 | — |
case-17 | fail→pass | 12,601 | 3,600 | -71% | 1 | 1 | 0% | 2,377 | 1,410 | -41% | 0 | 0 | — |
case-18 | fail→pass | 30,035 | 5,806 | -81% | 1 | 1 | 0% | 1,100 | 1,774 | +61% | 0 | 0 | — |
case-19 | fail→pass | 9,394 | 2,353 | -75% | 1 | 1 | 0% | 1,566 | 1,090 | -30% | 0 | 0 | — |
case-20 | fail→pass | 7,616 | 1,275 | -83% | 1 | 1 | 0% | 1,256 | 881 | -30% | 0 | 0 | — |
case-21 | pass→pass | 12,004 | 1,295 | -89% | 1 | 1 | 0% | 2,040 | 848 | -58% | 0 | 0 | — |
case-22 | fail→pass | 10,810 | 1,345 | -88% | 1 | 1 | 0% | 1,880 | 880 | -53% | 0 | 0 | — |
case-23 | pass→pass | 10,723 | 5,074 | -53% | 1 | 1 | 0% | 949 | 1,595 | +68% | 0 | 0 | — |
case-24 | pass→pass | 3,169 | 3,939 | +24% | 1 | 1 | 0% | 477 | 1,295 | +171% | 0 | 0 | — |
case-25 | pass→pass | 11,109 | 6,820 | -39% | 1 | 1 | 0% | 2,102 | 1,930 | -8% | 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 22 counted toward the lift figure. The other 3 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 +80 percentage points is the difference between those two pass rates over the 22 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.
Other measured skills in the registry, with their headline benchmark lift.