Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Add an agent to a workspace from the agent library. Proposes agent configuration including name, role, capabilities, adapter, and budget. Requires board approval via /approve before onboarding. Completes with identity setup and coordination prompt. Triggers on: "hire", "add agent", "onboard agent", "recruit"
.claude/skills/miosa-osa-hire/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 201% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 4% | 0% |
> Add an agent to the workspace from the library — with approval.
Bring a new agent into the workspace by selecting from the agent library, customizing its configuration, getting human approval, and onboarding it with identity and coordination protocols. This is the formal process for adding compute capacity — every hire goes through /approve to prevent uncontrolled agent sprawl and budget creep.
bash# Hire from library /hire researcher # Hire with customization /hire researcher --name "patent-analyst" --budget 80000 --adapter claude-opus # Browse available agents /hire --list # Hire with justification (for approval) /hire writer --reason "Need pitch deck content for board meeting Friday" # Hire multiple agents /hire researcher analyst --parallel # Preview what would be hired (no approval sent) /hire researcher --dry-run
| Flag | Type | Default | Description | |------|------|---------|-------------| | <agent> | positional | required | Agent type from library (or --list to browse) | | --name | string | from library | Custom name for this instance | | --role | string | from library | Custom role description | | --budget | int | from library | Token budget allocation | | --adapter | enum | from library | claude-opus, claude-sonnet, claude-haiku, codex, gemini, ollama | | --scope | string | from library | Workspace access scope | | --reason | string | — | Justification for hiring (included in approval request) | | --list | flag | false | Browse available agents in library | | --dry-run | flag | false | Preview configuration without creating approval | | --parallel | flag | false | Hire multiple agents simultaneously | | --skip-approval | flag | false | Skip approval (for pre-approved agent types) | | --tasks | string] | — | Initial tasks to assign after onboarding |
/approve as hire_agent type. Block until resolved. If rejected, report reason and stop. If revision requested, adjust and resubmit./budget/spawn. Deliver onboarding message: identity, available commands, scope boundaries, coordination protocols, initial tasks.markdown## Available Agents | Name | Role | Default Adapter | Default Budget | Capabilities | |------|------|-----------------|----------------|-------------| | researcher | Research Analyst | claude-sonnet | 50,000 | search, read, reduce, reflect | | writer | Content Writer | claude-sonnet | 30,000 | write, edit, translate | | coder | Software Engineer | claude-sonnet | 100,000 | read, write, build, test, debug | | analyst | Data Analyst | claude-sonnet | 50,000 | read, search, graph, stats | | reviewer | Code Reviewer | claude-sonnet | 20,000 | read, review, security-scan | | orchestrator | Team Lead | claude-opus | 150,000 | all coordination skills |
markdown## Agent Hired | Field | Value | |-------|-------| | Name | patent-analyst | | Role | Research Analyst (customized) | | Agent ID | agent-patent-analyst-a1b2 | | Adapter | claude-opus | | Budget | 80,000 tokens (~$12.00) | | Scope | node:miosa (read-only), processed/ (read-write) | | Approval | req-c3d4 (approved by Roberto) | | Status | onboarded, running | | Initial task | "Research AI patent landscape for MIOSA positioning" | Agent is live. Monitor with `/board` or message with `/inbox send patent-analyst "..."`.
/approve — Approval gate (required unless --skip-approval)/spawn — Agent runtime creation/budget — Budget allocation/inbox — Onboarding message delivery/heartbeat — Post-hire monitoring| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 18,722 | 7,732 | -59% | 1 | 1 | 0% | 3,415 | 2,224 | -35% | 0 | 0 | — |
case-02 | fail→fail | 4,868 | 9,008 | +85% | 1 | 1 | 0% | 779 | 2,707 | +247% | 0 | 0 | — |
case-03 | fail→fail | 21,766 | 8,840 | -59% | 1 | 1 | 0% | 2,340 | 2,456 | +5% | 0 | 0 | — |
case-04 | fail→pass | 9,015 | 3,936 | -56% | 1 | 1 | 0% | 1,590 | 1,775 | +12% | 0 | 0 | — |
case-05 | fail→fail | 22,908 | 16,001 | -30% | 1 | 1 | 0% | 1,009 | 2,238 | +122% | 0 | 0 | — |
case-14 | fail→fail | 14,530 | 2,167 | -85% | 1 | 1 | 0% | 2,458 | 1,434 | -42% | 0 | 0 | — |
case-06 | fail→fail | 10,198 | 5,111 | -50% | 1 | 1 | 0% | 1,542 | 2,038 | +32% | 0 | 0 | — |
case-07 | fail→pass | 5,749 | 7,979 | +39% | 1 | 1 | 0% | 823 | 2,476 | +201% | 0 | 0 | — |
case-08 | fail→pass | 12,983 | 2,561 | -80% | 1 | 1 | 0% | 1,995 | 1,658 | -17% | 0 | 0 | — |
case-09 | fail→pass | 21,256 | 4,319 | -80% | 1 | 1 | 0% | 3,313 | 1,954 | -41% | 0 | 0 | — |
case-10 | fail→pass | 10,960 | 4,669 | -57% | 1 | 1 | 0% | 1,824 | 1,896 | +4% | 0 | 0 | — |
case-11 | fail→pass | 6,380 | 7,600 | +19% | 1 | 1 | 0% | 810 | 2,549 | +215% | 0 | 0 | — |
case-12 | fail→pass | 9,857 | 1,962 | -80% | 1 | 1 | 0% | 1,635 | 1,436 | -12% | 0 | 0 | — |
case-13 | pass→pass | 7,735 | 2,952 | -62% | 1 | 1 | 0% | 1,222 | 1,770 | +45% | 0 | 0 | — |
case-15 | fail→pass | 26,781 | 2,636 | -90% | 1 | 1 | 0% | 1,562 | 1,492 | -4% | 0 | 0 | — |
case-16 | fail→fail | 13,022 | 1,881 | -86% | 1 | 1 | 0% | 1,835 | 1,430 | -22% | 0 | 0 | — |
case-17 | fail→fail | 11,095 | 2,318 | -79% | 1 | 1 | 0% | 1,495 | 1,492 | -0% | 0 | 0 | — |
case-18 | pass→pass | 16,930 | 3,518 | -79% | 1 | 1 | 0% | 2,859 | 1,542 | -46% | 0 | 0 | — |
case-19 | fail→pass | 8,995 | 6,248 | -31% | 1 | 1 | 0% | 1,481 | 2,299 | +55% | 0 | 0 | — |
case-20 | fail→pass | 4,942 | 10,461 | +112% | 1 | 1 | 0% | 627 | 3,015 | +381% | 0 | 0 | — |
case-21 | fail→pass | 4,732 | 3,351 | -29% | 1 | 1 | 0% | 676 | 1,535 | +127% | 0 | 0 | — |
case-22 | fail→pass | 9,141 | 10,984 | +20% | 1 | 1 | 0% | 1,094 | 2,761 | +152% | 0 | 0 | — |
case-23 | pass→pass | 8,979 | 3,107 | -65% | 1 | 1 | 0% | 1,319 | 1,679 | +27% | 0 | 0 | — |
case-24 | fail→pass | 15,315 | 2,215 | -86% | 1 | 1 | 0% | 2,341 | 1,424 | -39% | 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. 24 cases were attempted, and 23 counted toward the lift figure. The other 1 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 +54 percentage points is the difference between those two pass rates over the 23 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.