Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use Orca orchestration for structured multi-agent coordination: threaded messages, blocking ask/reply flows, task dispatch, worker_done/escalation waits, task DAGs, decision gates, coordinator loops, or decomposing work across agents. Use `orca-cli` instead for full ownership handoffs, including requests phrased as "hand off", "handoff", "handover", "give this to another agent", or "another worktree" when the user did not explicitly ask to supervise, monitor, wait for results, or coordinate a DA
.claude/skills/stablyai-orchestration/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 44% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 51% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -69% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -23% | 0% |
This file is a discovery stub, not the usage guide. The full, version-matched Orca orchestration reference is served by the orca binary itself — kept out of this file on purpose so it can never drift from the binary that will actually run your commands.
Engage Orca orchestration whenever you need structured multi-agent coordination: threaded messages, blocking ask/reply flows, task dispatch, worker_done/escalation waits, task DAGs, decision gates, coordinator loops, or decomposing work across agents. Use the orca-cli skill instead for full ownership handoffs ("hand off", "handoff", "handover", "give this to another agent", "another worktree") when the user did not ask to supervise, monitor, wait for results, or coordinate a DAG — and for ordinary terminal control, shell commands, worktree management, and the built-in browser. Coordination requires real Orca runtime state; never substitute a non-Orca subagent tool.
Choose the executable once and reuse it for every later command:
ORCA_CLI_COMMAND environment variable is set, use its value. Orca exports thisfor managed WSL sessions.
ORCA_DEV_REPO_ROOT, use orca-dev.orca-ide. Never run bareorca there — outside Orca's terminals it normally resolves to the GNOME Orca screen reader (/usr/bin/orca) and starts speech on the user's machine.
orca.Below, ORCA is a placeholder for the executable you resolved. Substitute it before running anything; do not create a shell variable or run ORCA literally. This works the same way in POSIX shells, PowerShell, and cmd.exe.
If the selected executable cannot run, report its exact error and stop. Do not fall through to another executable, which could silently target a different Orca build.
textORCA skills get orchestration
That prints the compact, version-matched guide for the exact binary that will handle your next commands. It covers the normal local coordinator loop. For a conditional action gate such as remote placement, uncertain release recovery, or expanded DAG work, load only the reference that gate names with ORCA skills get orchestration --reference references/<file>.md (--references lists the names). If that binary rejects --reference, run ORCA skills get orchestration --full and read the named bundled reference before acting.
Prefer --json. Use the selected executable's --help for commands or flags the guide does not cover. If a command reports that Orca is not running, start it with ORCA open --json and retry. If skills get is unknown, explain that updating Orca restores the guide; use --help for read-only discovery and do not guess unsupported commands.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 24,521 | 17,887 | -27% | 1 | 1 | 0% | 3,512 | 1,233 | -65% | 0 | 0 | — |
case-16 | pass→pass | 20,226 | 28,804 | +42% | 1 | 1 | 0% | 1,203 | 2,881 | +139% | 0 | 0 | — |
case-02 | fail→fail | 29,232 | 17,822 | -39% | 1 | 1 | 0% | 3,274 | 1,217 | -63% | 0 | 0 | — |
case-03 | fail→fail | 29,068 | 17,358 | -40% | 1 | 1 | 0% | 5,011 | 1,221 | -76% | 0 | 0 | — |
case-04 | fail→pass | 12,501 | 18,952 | +52% | 1 | 1 | 0% | 1,103 | 1,586 | +44% | 0 | 0 | — |
case-05 | fail→pass | 9,398 | 7,667 | -18% | 1 | 1 | 0% | 699 | 1,055 | +51% | 0 | 0 | — |
case-15 | pass→pass | 13,221 | 9,818 | -26% | 1 | 1 | 0% | 1,396 | 1,487 | +7% | 0 | 0 | — |
case-06 | fail→pass | 11,969 | 9,224 | -23% | 1 | 1 | 0% | 1,082 | 1,466 | +35% | 0 | 0 | — |
case-07 | pass→pass | 27,681 | 7,974 | -71% | 1 | 1 | 0% | 1,324 | 1,196 | -10% | 0 | 0 | — |
case-08 | pass→pass | 16,915 | 8,783 | -48% | 1 | 1 | 0% | 1,797 | 1,306 | -27% | 0 | 0 | — |
case-09 | fail→pass | 32,439 | 7,472 | -77% | 1 | 1 | 0% | 3,281 | 1,007 | -69% | 0 | 0 | — |
case-10 | fail→pass | 12,516 | 7,118 | -43% | 1 | 1 | 0% | 1,305 | 1,007 | -23% | 0 | 0 | — |
case-11 | fail→pass | 14,386 | 8,780 | -39% | 1 | 1 | 0% | 1,582 | 1,314 | -17% | 0 | 0 | — |
case-12 | fail→fail | 13,706 | 7,243 | -47% | 1 | 1 | 0% | 1,277 | 871 | -32% | 0 | 0 | — |
case-13 | fail→fail | 13,772 | 7,912 | -43% | 1 | 1 | 0% | 1,373 | 1,126 | -18% | 0 | 0 | — |
case-14 | pass→pass | 14,749 | 8,278 | -44% | 1 | 1 | 0% | 1,581 | 1,217 | -23% | 0 | 0 | — |
case-17 | pass→fail | 9,557 | 35,183 | +268% | 1 | 1 | 0% | 736 | 3,142 | +327% | 0 | 0 | — |
case-18 | fail→pass | 16,949 | 52,575 | +210% | 1 | 1 | 0% | 283 | 8,404 | +2870% | 0 | 0 | — |
case-19 | fail→pass | 20,732 | 13,533 | -35% | 1 | 1 | 0% | 2,798 | 2,237 | -20% | 0 | 0 | — |
case-20 | pass→pass | 9,972 | 7,475 | -25% | 1 | 1 | 0% | 639 | 1,016 | +59% | 0 | 0 | — |
case-21 | pass→pass | 22,760 | 7,233 | -68% | 1 | 1 | 0% | 2,796 | 1,007 | -64% | 0 | 0 | — |
case-22 | fail→pass | 22,799 | 13,043 | -43% | 1 | 1 | 0% | 2,939 | 2,404 | -18% | 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 +36 percentage points is the difference between those two pass rates over the 17 comparable cases. 1 case got worse with the skill loaded, and it is 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/7/2026 | +43% |
Other measured skills in the registry, with their headline benchmark lift.