メニュー
Execute Python in Daytona sandboxes with durable volume persistence. Use when running code, persisting results, inspecting volumes, managing workspace lifecycle, or validating Daytona connectivity.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Delegate recursive work to child RLM sandboxes with budget management. Use when decomposing tasks into sub-queries, fanning out batched work, managing LLM call budgets, or building parent-child RLM hierarchies.
Diagnose fleet-rlm runtime failures, API contract drift, sandbox errors, and observability issues. Use when something is broken — Daytona connection failures, websocket mismatches, escalation not triggering, budget exhaustion, or missing traces.
Optimize fleet-rlm DSPy programs and RLM skill bundles using GEPA with MLflow tracking. Use when running optimization loops, designing feedback metrics, building training datasets, or comparing runs.
Hub skill for fleet-rlm: when to use dspy.RLM vs ReAct/CodeAct and which workflow skill to load next.
Design DSPy signatures and compose runtime modules for fleet-rlm tasks. Use when creating input/output field definitions, choosing between built-in signatures, selecting execution modes, or wiring custom modules.
Fetch and inspect JavaScript-heavy pages with Playwright in a Daytona browser-capable snapshot.
| name | sandbox-execution |
| description | Execute Python in Daytona sandboxes with durable volume persistence. Use when running code, persisting results, inspecting volumes, managing workspace lifecycle, or validating Daytona connectivity. |
| Parameter | Default | Description |
|---|---|---|
timeout | 900 | Sandbox lifetime seconds |
execute_timeout | None | Per-execute timeout, defaults to timeout |
volume_name | None | Durable volume name for cross-session persistence |
repo_url | None | Repo to stage into sandbox |
repo_ref | None | Branch/commit for staging |
context_paths | None | Specific paths to stage |
max_llm_calls | 50 | Semantic LLM budget |
llm_call_timeout | 60 | Per LLM call timeout |
async_execute | True | Use async execution path |
max_recursion_depth | 2 | Delegation depth limit |
delegate_max_iterations | 8 | Per-child RLM iterations |
SUBMIT(field=value, ...) to return results.result.field_name (NOT result["field"])./home/daytona/memory/.references/volume-layout.md for full directory table.add_buffer(name, value) / get_buffer(name) — injected by sandbox driver.execute() calls within same session.uv run fleet-rlm daytona-smoke --repo <url> [--ref <branch>]
Run before depending on live Daytona behavior.
from fleet_rlm.integrations.daytona.interpreter import DaytonaInterpreter
interpreter = DaytonaInterpreter(timeout=300)
try:
await interpreter.start()
result = await interpreter.execute("""
import math
SUBMIT(answer=math.factorial(10), status="ok")
""")
print(result.answer) # 3628800
finally:
await interpreter.shutdown()
interpreter = DaytonaInterpreter(timeout=300, volume_name="my-project")
try:
await interpreter.start()
# Write to durable volume
await interpreter.execute("""
import json, pathlib
out = pathlib.Path("/home/daytona/memory/knowledge/report.json")
out.parent.mkdir(parents=True, exist_ok=True)
out.write_text(json.dumps({"key": "value"}))
SUBMIT(saved=str(out))
""")
# Read back in a later call (or even a later session with same volume_name)
result = await interpreter.execute("""
import json, pathlib
data = json.loads(pathlib.Path("/home/daytona/memory/knowledge/report.json").read_text())
SUBMIT(data=data)
""")
print(result.data) # {"key": "value"}
finally:
await interpreter.shutdown()
interpreter = DaytonaInterpreter(timeout=600)
try:
await interpreter.start()
# Step 1: collect items
await interpreter.execute("""
add_buffer("findings", "First observation")
add_buffer("findings", "Second observation")
SUBMIT(step="collected")
""")
# Step 2: aggregate
result = await interpreter.execute("""
items = get_buffer("findings")
SUBMIT(count=len(items), items=items)
""")
print(result.items) # ["First observation", "Second observation"]
finally:
await interpreter.shutdown()