Install any skill in seconds. Free to start, no credit card required.
Get Started Free →One-command workspace activation from a template. Reads a TOML/YAML workspace template, spawns agents, assigns initial tasks, sets budgets, and bootstraps the full operating environment. Variable substitution for customization. Triggers on: "launch", "start workspace", "activate", "boot workspace"
.claude/skills/miosa-osa-launch/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 1317% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 1546% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -65% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -61% | 0% |
> One-command workspace activation from template.
Stand up a fully operational workspace in a single command. A workspace template defines everything: agents to hire, skills to enable, initial tasks, budget allocations, file structure, and configuration. Launch reads the template, provisions all resources, and starts execution. Variable substitution lets you customize templates for different projects without editing them.
bash# Launch from a template /launch templates/research-team.yaml # Launch with variable substitution /launch templates/project.yaml --var project_name="AI Masters" --var budget=100000 # Launch from the template library /launch --template content-pipeline # Preview without launching /launch templates/research-team.yaml --dry-run # Launch with custom budget override /launch templates/research-team.yaml --budget 200000 # Launch and assign initial task /launch templates/research-team.yaml --task "Research competitor pricing"
| Flag | Type | Default | Description | |------|------|---------|-------------| | <template> | positional | required | Path to template file or library name | | --template | string | — | Template name from the library | | --var | key=value] | — | Variable substitution (overrides template defaults) | | --budget | int | from template | Override total budget | | --task | string | from template | Override or set initial task | | --dry-run | flag | false | Preview what would be created | | --skip-approval | flag | false | Skip hire approvals (for trusted templates) | | --backend | enum | from template | Override agent backend: subprocess, tmux, docker | | --output-dir | path | . | Workspace root directory |
{{variable}} placeholders with values from --var flags or template defaults.agents/, inbox/, processed/, ops/, .locks/, approvals/./hire (with or without approval based on --skip-approval).markdown## Workspace Launched **Template:** research-team **Directory:** /workspace/ai-masters-research **Budget:** 200,000 tokens ($15.00) ### Structure Created
ai-masters-research/ ├── SYSTEM.md ├── agents/ │ ├── researcher/ │ ├── analyst/ │ └── writer/ ├── inbox/ ├── processed/ ├── ops/ │ ├── observations/ │ ├── metrics/ │ └── budget/ ├── .locks/ └── approvals/
### Agents Hired
| Agent | Role | Adapter | Budget | Status |
|-------|------|---------|--------|--------|
| researcher | Research Analyst | claude-sonnet | 80,000 | running |
| analyst | Data Analyst | claude-sonnet | 60,000 | running |
| writer | Content Writer | claude-haiku | 30,000 | running |
| unallocated | — | — | 30,000 | reserve |
### Initial Tasks
| Task | Assigned To | Priority |
|------|-------------|----------|
| Research competitor pricing | researcher | high |
| Analyze existing revenue data | analyst | high |
| Draft findings report | writer | normal (after research) |
### Ready
Workspace is live. Monitor with `/board`. Send messages with `/inbox`./hire — Agent provisioning/budget — Budget initialization/spawn — Agent runtime creation/heartbeat — Monitoring setup/board — Dashboard activation| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 26,584 | 7,899 | -70% | 1 | 1 | 0% | 4,140 | 2,489 | -40% | 0 | 0 | — |
case-02 | fail→pass | 16,630 | 7,059 | -58% | 1 | 1 | 0% | 2,360 | 2,390 | +1% | 0 | 0 | — |
case-03 | fail→fail | 5,610 | 7,954 | +42% | 1 | 1 | 0% | 283 | 2,658 | +839% | 0 | 0 | — |
case-04 | fail→fail | 12,197 | 22,211 | +82% | 1 | 1 | 0% | 915 | 4,013 | +339% | 0 | 0 | — |
case-05 | fail→pass | 5,544 | 9,121 | +65% | 1 | 1 | 0% | 187 | 2,650 | +1317% | 0 | 0 | — |
case-06 | fail→pass | 5,887 | 20,498 | +248% | 1 | 1 | 0% | 249 | 4,099 | +1546% | 0 | 0 | — |
case-07 | fail→pass | 23,569 | 2,271 | -90% | 1 | 1 | 0% | 3,933 | 1,388 | -65% | 0 | 0 | — |
case-08 | fail→pass | 22,171 | 3,088 | -86% | 1 | 1 | 0% | 3,627 | 1,403 | -61% | 0 | 0 | — |
case-09 | fail→pass | 9,238 | 2,382 | -74% | 1 | 1 | 0% | 1,691 | 1,348 | -20% | 0 | 0 | — |
case-10 | pass→pass | 18,509 | 15,977 | -14% | 1 | 1 | 0% | 2,652 | 3,636 | +37% | 0 | 0 | — |
case-15 | fail→pass | 9,804 | 2,030 | -79% | 1 | 1 | 0% | 1,596 | 1,347 | -16% | 0 | 0 | — |
case-11 | fail→pass | 18,352 | 6,153 | -66% | 1 | 1 | 0% | 2,428 | 1,973 | -19% | 0 | 0 | — |
case-12 | fail→pass | 11,824 | 3,791 | -68% | 1 | 1 | 0% | 1,980 | 1,389 | -30% | 0 | 0 | — |
case-13 | pass→pass | 15,872 | 5,184 | -67% | 1 | 1 | 0% | 1,548 | 1,749 | +13% | 0 | 0 | — |
case-14 | fail→pass | 10,427 | 3,134 | -70% | 1 | 1 | 0% | 2,018 | 1,459 | -28% | 0 | 0 | — |
case-21 | fail→pass | 15,541 | 4,539 | -71% | 1 | 1 | 0% | 2,715 | 1,914 | -30% | 0 | 0 | — |
case-16 | pass→pass | 8,132 | 2,782 | -66% | 1 | 1 | 0% | 1,305 | 1,429 | +10% | 0 | 0 | — |
case-17 | fail→pass | 5,219 | 4,003 | -23% | 1 | 1 | 0% | 208 | 1,811 | +771% | 0 | 0 | — |
case-18 | fail→pass | 14,609 | 2,124 | -85% | 1 | 1 | 0% | 1,870 | 1,252 | -33% | 0 | 0 | — |
case-19 | fail→pass | 11,702 | 2,523 | -78% | 1 | 1 | 0% | 1,829 | 1,501 | -18% | 0 | 0 | — |
case-20 | fail→pass | 16,211 | 2,500 | -85% | 1 | 1 | 0% | 2,155 | 1,315 | -39% | 0 | 0 | — |
case-22 | pass→pass | 5,487 | 2,030 | -63% | 1 | 1 | 0% | 856 | 1,388 | +62% | 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 17 counted toward the lift figure. The other 5 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 +68 percentage points is the difference between those two pass rates over the 17 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.