Menú
Interpret Jupyter execution sessions in isanagent—attachments, kernel cwd, tokens; contrast with local provider behavior.
Instalar con Codex o Claude Copia este prompt, pégalo en Codex, Claude u otro asistente, y deja que revise la página de la skill y la instale por ti.
Basado en la clasificación ocupacional SOC
MaxEvolve Triton/PyTorch to JAX/Pallas porting with GpuToJax 12-step pipeline, interpret=True validation, MAP-Elites evolution, and Colab/SSH hardware profiling.
Manage Google Colab sessions and execute code on remote Colab VMs via the colab-cli.
Use when you need to generate a synthetic dataset. You can use Afterimage Python package to create, conversational, tool-calling, multi-choice questions, or other types of structured output datasets for SFT, or preference data for DPO.
Schedule deferred or recurring work via the isanagent cron tool—add/remove jobs with correct runtime context fields.
Configure long wall clocks and use execution_run_background + job polling for ML or hours-long runs without blocking the agent loop.
Run numeric or ML code through isanagent execution tools without flooding context—artifacts, incremental runs, sandbox paths.
| name | jupyter-heavy-output |
| description | Interpret Jupyter execution sessions in isanagent—attachments, kernel cwd, tokens; contrast with local provider behavior. |
| requires | {"bins":[],"env":[]} |
| always | false |
Apply this when execution_env_info (or the session you opened) shows the Jupyter provider—i.e. the host configured default_provider = "jupyter" and valid [harness.execution.jupyter] settings.
execution_session_create returns session_capabilities (JSON) for that session. Jupyter sessions use a remote kernel with persistent variables across execution_run calls until the session is closed or the kernel dies.execution_run may return attachments: sandbox-relative paths for materialized image/png, image/jpeg, large text/csv, large application/json, etc. Short human-readable lines may appear in stdout; do not assume full binary or huge JSON appears in the tool string—open paths via execution_artifact_list + file tools, and describe outcomes in prose.execution_run only supports cwd_mode: session_default in practice—the process cwd is on the Jupyter server, not your agent sandbox. If the user needs a server-side directory change, encode it in code (e.g. os.chdir / %cd) explicitly; do not confuse server paths with sandbox_relative cwd for the local provider.| Concern | Local provider | Jupyter provider |
|---|---|---|
| Where code runs | Child / REPL worker on agent host, cwd under sandbox | Remote kernel on Jupyter server |
| Packages | Same as agent host python | Whatever is installed on the server |
| Big plot/table | Prefer write under .execution_artifacts/... or attachments pattern | Often attachments from display_data |
| Interrupt | execution_cancel + harness behavior | Uses server interrupt when supported |
Check execution_session_create metadata (e.g. extensions.local_python_repl on local) when you need to remember whether local Python is REPL-style or subprocess—do not guess from memory.
Prefer the JUPYTER_TOKEN environment variable on the agent process over pasting tokens into replies. Never echo tokens into chat or logs.